Encode ibm32 header fields on write

[andgineer]

Problem

SegyFactory.create_traces does not IBM-encode ibm32 header fields, even though TraceAccessor IBM-decodes them on read. The write path is therefore asymmetric with the read path: an ibm32 header written by the factory is stored as a raw integer, and reading it back through the IBM decode yields garbage.

We hit this in a production MDIO->SEG-Y cut: an ibm32 header (azimuth) came out of the cut as a plain integer, which QC then flagged because the value no longer matched the source.

Fix

• create_traces: add a keyed to_ibm transform to the header pipeline, placed before the byte-swap. This is the exact inverse of the read path (byte_swap then to_ieee) and mirrors the data pipeline's existing ibm32 handling.
• create_trace_header_template: expose ibm32 fields as float32 instead of uint32, so callers fill real float values that then get encoded. This matches what the reader returns for ibm32 headers and what create_trace_sample_template already does for ibm32 samples.

The on-disk format is unchanged (4 bytes, IBM-encoded, big-endian); only the in-memory template dtype changes.

Open question

The create_traces encode change fixes the bug on its own, because the reader already hands callers a float32 array.

The create_trace_header_template -> float32 change is the consistency/usability part, and it's the only piece that changes a public dtype. Happy to drop it and keep the PR to just the encode if you'd rather avoid the template dtype change. Let me know which you prefer.

Tests

• New TestSegyFactoryIbmHeader (big/little x 1/42 traces): on-disk bytes equal ieee2ibm(value), plus a lossless round-trip.
• Full suite: 289 passed. mypy and ruff clean.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.94%. Comparing base (818c1a5) to head (be68c2e).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #376      +/-   ##
==========================================
+ Coverage   95.86%   95.94%   +0.07%     
==========================================
  Files          50       50              
  Lines        2663     2713      +50     
  Branches      149      151       +2     
==========================================
+ Hits         2553     2603      +50     
  Misses         84       84              
  Partials       26       26              

☔ View full report in Codecov by Harness.
📢 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.

[BrianMichell]

The create_trace_header_template -> float32 change is the consistency/usability part

I think it's appropriate for the bugfix to ensure we don't cause inadvertant trucation. float32 dtype is the way to go.

[BrianMichell]

Please ensure that the codecov pipeline passes. I think applying that one change should satisfy it.

[andgineer]

Thanks! The PR already has the float32 version, so nothing to change - good to merge whenever.