subsetdims: support full coordinate permutations (p == d)

[thisiscam]

subsetdims(C, dims) currently asserts p < d, so a dims that is a non-identity permutation of all d coordinates throws — even though it is a well-defined reordering. (This came up in the #367 review, re: the p == d reordering case.)

This relaxes the assert to p <= d. The identity dims == 1:d still short-circuits to C, and p == 1 to a Uniform, so the only newly-allowed case is a genuine permutation. Invalid inputs (p > d, duplicate dims, out-of-range dims) are still rejected by the remaining asserts.

The generic SubsetCopula machinery and the per-family specializations already handle p == d:

• elliptical (Gaussian/T): reindex Σ[dims, dims] — respects the permutation;
• Archimedean: exchangeable, so any permutation yields the same copula;
• generic: _cdf saturates 0 dims (just permutes), _logpdf reduces to the full density, _rand! permutes a parent draw.

Adds test/Subsetting.jl: the regression (p == d no longer throws) plus a correctness check on an asymmetric Gaussian where the permutation is non-trivial (subsetdims(C, perm).Σ == Σ[perm, perm], and cdf/logpdf match the permuted parent).

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 78.28%. Comparing base (5cd3d24) to head (a56e231).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #369   +/-   ##
=======================================
  Coverage   78.28%   78.28%           
=======================================
  Files          84       84           
  Lines        5106     5106           
=======================================
  Hits         3997     3997           
  Misses       1109     1109           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[lrnv]

Work for me, thanks for the catch :)