Test: reproduce/guard data.tar inflation over the archive lifecycle

[egpbos]

Description

Phase 1 of an investigation into data.tar archive inflation. Production profiling of a 1.1 TiB, 54-run dataset found ~97–98% of data.tar bytes were byte-identical duplicates. This PR pins the behaviour with tests before any redesign (Phase 2, separate); no production code is changed.

Adds two end-to-end tests to tests/utils/test_archive.py that replay the full proteus.py archive lifecycle (rolling update + remove_old cycles, then the end-of-run update(remove_files=True) full archive) on a realistic mini output/data/, and measure inflation by member content (SHA-256), not just by name:

• test_lifecycle_final_pass_stores_reference_data_once — regression guard that the static EOS table (spider_eos/) and fixed-name runtime files enter data.tar exactly once across the whole lifecycle. Extends the rolling-only coverage of test_update_snapshots_only_bounds_tar_growth_across_cycles (Super-Earth structure and thermal evolution: interior coupling, molten initial conditions #706) to the final remove_files=True pass, and adds a content-level check that a fixed-name file which legitimately changes each cycle (zalmoxis_output.dat) is preserved as distinct versions rather than mislabelled as duplication.

• test_lifecycle_snapshot_reappend_no_inflation — encodes the target contract (store each snapshot once, inflation ≈ 1.0). Marked xfail(strict=True): on main every timestamped snapshot lands in data.tar ~2×, because the rolling archive keeps snapshots loose (remove_files=False) so each is re-appended on the next cycle before remove_old prunes it, and once more by the final full-archive pass (observed inflation ~1.45× on this fixture). strict so it flips to a hard failure — forcing the marker's removal — once the archive redesign removes the duplication.

Also adds a _tar_content_stats helper (gross/distinct bytes, per-name member and distinct-content counts). Pre-existing archive tests are unchanged.

Verdict recorded (measured on this fixture, not assumed): EOS/fixed-name files are fixed on main by #706's snapshots_only=True guard (green guard); .nc/.json snapshots still inflate on main (~2× per snapshot, strict xfail). The ~1.45× figure is fixture-specific and is not a claim about the production ratio.

No linked issue yet — Phase 1 of the archiving-inflation investigation.

Validation of changes

• pytest tests/utils/test_archive.py -q → 21 passed, 1 xfailed, no regressions in the pre-existing archive tests.
• pytest ... --runxfail confirms the xfail is a genuine assert 2 == 1 (snapshot stored twice), not an incidental error masquerading as expected-fail.
• Quality/lint gates clean: tools/check_test_quality.py --check, ruff check/format, tools/validate_test_structure.sh.
• Test configuration: macOS (Darwin) with Python 3.12; CI additionally runs Unit Tests on Linux and macOS — all green on this PR.

Open question for reviewers (feeds the redesign, not asserted here): the snapshots_only=False path also retains multiple distinct versions of zalmoxis_output.dat (append-only tar keeps every version under one name). Before deciding whether a redesign must preserve that version history, confirm whether any resume/extract path ever reads an older in-tar version or whether an overwrite was intended (tarfile.extractall lets the last member win).

Checklist

• I have followed the contributing guidelines
• My code follows the style guidelines of this project
• I have performed a self-review of my code
• My changes generate no new warnings or errors
• I have checked that the tests still pass on my computer
• I have updated the docs, as appropriate (n/a — test-only change)
• I have added tests for these changes, as appropriate
• I have checked that all dependencies have been updated, as required (n/a — no dependency changes)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.67%. Comparing base (9881061) to head (c482dc0).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #781   +/-   ##
=======================================
  Coverage   92.66%   92.67%           
=======================================
  Files         110      110           
  Lines       15807    15807           
  Branches     2846     2846           
=======================================
+ Hits        14648    14649    +1     
+ Misses       1159     1158    -1     

Flag	Coverage Δ
unit-tests	83.64% <ø> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ 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.

[timlichtenberg]

@egpbos Please can you use the PROTEUS PR templates when creating the PRs?

[egpbos]

@egpbos Please can you use the PROTEUS PR templates when creating the PRs?

Done.