docs: add summary lines to 19 reducer operation docstrings

[TaiSakuma]

Summary

Rolls the pilot from #3946 out to the 19 operations (28 functions, including the nan* variants) in the Reducers batch of #3980:

all, any, sum, prod, count, count_nonzero, min, max, argmin, argmax, mean, std, var, moment, corr, covar, ptp, linear_fit, softmax.

Each docstring now has:

• A one-line summary on the opening """.
• Returns: and Examples: section headers (the latter only where examples exist).

Original body text is preserved; only structural edits are applied. The summary lines were reviewed against the checklist posted in #3980.

Summary lines

Operation	Summary
ak.all	Returns whether all elements are True over one or all levels of nesting.
ak.any	Returns whether any elements are True over one or all levels of nesting.
ak.sum	Sums an array's elements over one or all levels of nesting.
ak.nansum	Sums an array's elements, treating NaN values as missing.
ak.prod	Multiplies an array's elements over one or all levels of nesting.
ak.nanprod	Multiplies an array's elements, treating NaN values as missing.
ak.count	Counts an array's elements over one or all levels of nesting.
ak.count_nonzero	Counts an array's nonzero elements over one or all levels of nesting.
ak.min	Returns the minimum value over one or all levels of nesting.
ak.nanmin	Returns the minimum value, treating NaN values as missing.
ak.max	Returns the maximum value over one or all levels of nesting.
ak.nanmax	Returns the maximum value, treating NaN values as missing.
ak.argmin	Returns the index of the minimum value over one or all levels of nesting.
ak.nanargmin	Returns the index of the minimum value, treating NaN values as missing.
ak.argmax	Returns the index of the maximum value over one or all levels of nesting.
ak.nanargmax	Returns the index of the maximum value, treating NaN values as missing.
ak.mean	Computes the mean over one or all levels of nesting.
ak.nanmean	Computes the mean, treating NaN values as missing.
ak.std	Computes the standard deviation over one or all levels of nesting.
ak.nanstd	Computes the standard deviation, treating NaN values as missing.
ak.var	Computes the variance over one or all levels of nesting.
ak.nanvar	Computes the variance, treating NaN values as missing.
ak.moment	Computes the nth moment over one or all levels of nesting.
ak.corr	Computes the correlation of x and y over one or all levels of nesting.
ak.covar	Computes the covariance of x and y over one or all levels of nesting.
ak.ptp	Returns the range of values over one or all levels of nesting.
ak.linear_fit	Computes the linear fit of y against x over one or all levels of nesting.
ak.softmax	Computes the softmax over the innermost level of nesting.

Notes for reviewers

• Base reducers share the tail "…over one or all levels of nesting." to cover both axis=None (scalar result) and axis=int (group-wise); nan* variants share "…, treating NaN values as missing." and keep the verbatim "Like #ak.X, but…" text in the body.
• ak.softmax deviates deliberately ("over the innermost level of nesting"): its implementation only supports the innermost axis and raises NotImplementedError otherwise.
• ak.argmin/ak.argmax read "Returns the index of the minimum/maximum value…"; the "…, returning integer indexes" register of ak.argcartesian did not fit the line budget with the modes tail.
• Pre-existing issues left verbatim in the bodies (out of scope here): nanargmin's Args block documents mask_identity twice, and ak.mean's first walkthrough doctest is malformed in main (missing ... continuation prefixes).

Test plan

• pre-commit passes locally on the modified files (ruff check, ruff format, codespell, mypy).
• Mechanical validation: code outside docstrings byte-identical; Args: blocks and doctests byte-identical; no original narrative text lost.
• Sphinx docs build cleanly on CI.
• Rendered docstrings spot-checked in the docs preview.

Part of #3980. Draft for initial review; will mark ready once CI passes and the preview is checked.

AI assistance disclosure

Drafted with Claude Code: summaries drafted by Claude Opus 4.8, adversarially verified against the implementations by independent Claude Opus 4.8 agents, mechanically validated (docstring-only changes, original text preserved), and reviewed by a human before submission.

Co-Authored-By: Claude Opus 4.8 noreply@anthropic.com

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 84.03%. Comparing base (c4aab45) to head (9289ef1).

Additional details and impacted files

Files with missing lines	Coverage Δ
src/awkward/operations/ak_all.py	96.66% <ø> (ø)
src/awkward/operations/ak_any.py	96.66% <ø> (ø)
src/awkward/operations/ak_argmax.py	94.59% <ø> (ø)
src/awkward/operations/ak_argmin.py	94.59% <ø> (ø)
src/awkward/operations/ak_corr.py	97.43% <ø> (ø)
src/awkward/operations/ak_count.py	100.00% <ø> (ø)
src/awkward/operations/ak_count_nonzero.py	96.55% <ø> (ø)
src/awkward/operations/ak_covar.py	97.05% <ø> (ø)
src/awkward/operations/ak_linear_fit.py	100.00% <ø> (ø)
src/awkward/operations/ak_max.py	94.73% <ø> (ø)
... and 9 more

[TaiSakuma]

Follow-up per @ikrommyd's review on #4130: the Returns: sections now describe the returned object (a noun phrase) rather than restating the operation. Final Returns: lines for this batch:

Returns

Operation	Returns
ak.all	True in each group of elements from array (many types supported, including all Awkward Arrays and Records) if all values are True; False otherwise.
ak.any	True in each group of elements from array (many types supported, including all Awkward Arrays and Records) if any values are True; False otherwise.
ak.sum	The sum over array (many types supported, including all Awkward Arrays and Records).
ak.prod	The product of the elements of array (many types supported, including all Awkward Arrays and Records).
ak.count	The number of elements of array (many types supported, including all Awkward Arrays and Records).
ak.count_nonzero	The number of nonzero elements of array (many types supported, including all Awkward Arrays and Records).
ak.min	The minimum value in each group of elements from array (many types supported, including all Awkward Arrays and Records).
ak.max	The maximum value in each group of elements from array (many types supported, including all Awkward Arrays and Records).
ak.argmin	The index position of the minimum value in each group of elements from array (many types supported, including all Awkward Arrays and Records).
ak.argmax	The index position of the maximum value in each group of elements from array (many types supported, including all Awkward Arrays and Records).
ak.mean	The mean in each group of elements from x (many types supported, including all Awkward Arrays and Records).
ak.std	The standard deviation in each group of elements from x (many types supported, including all Awkward Arrays and Records).
ak.var	The variance in each group of elements from x (many types supported, including all Awkward Arrays and Records).
ak.moment	The nth moment in each group of elements from x (many types supported, including all Awkward Arrays and Records).
ak.corr	The correlation of x and y (many types supported, including all Awkward Arrays and Records, must be broadcastable to each other).
ak.covar	The covariance of x and y (many types supported, including all Awkward Arrays and Records, must be broadcastable to each other).
ak.ptp	The range of values in each group of elements from array (many types supported, including all Awkward Arrays and Records).
ak.linear_fit	The linear fit of y with respect to x (many types supported, including all Awkward Arrays and Records, must be broadcastable to each other).
ak.softmax	The softmax in each group of elements from x (many types supported, including all Awkward Arrays and Records).

[github-actions]

The documentation preview is ready to be viewed at http://preview.awkward-array.org.s3-website.us-east-1.amazonaws.com/PR4129

[ianna]

@TaiSakuma - this looks great! Thanks.