fix(plan): do not use min/max stats for >= NaN float predicate

[jonathansergio]

Resolves #26805

What does this PR do?

Fixes an incorrect skip-batch optimization for floating-point columns: prevent using min/max statistics when the operator is >= and the literal is NaN.

Why?

Min/max stats exclude NaN. For col >= NaN, NaN satisfies the predicate (TotalOrd: NaN >= NaN is true). Generating a skip predicate like max(A) < NaN may evaluate to true (e.g. 0.0 < NaN) and incorrectly skip a batch that contains NaN, producing wrong results.

Changes

• Updated can_use_min_max_stats in skip_batches.rs:
  • Split Gt and GtEq handling:
  • Keep Some(O::Gt) => lv_is_nan (safe: nothing is greater than NaN).
  • Change Some(O::GtEq) => false (unsafe: >= NaN may match NaN rows but stats exclude NaN).
• Expanded docstring to explain rationale.
• Tests in test_skip_batch_predicate.py:
  • Fixed an incorrect assertion for >= NaN.
  • Added a deterministic regression test test_float_geq_nan_skip_batch_predicate() that ensures col >= NaN does not produce a skip predicate and that rows with NaN are not skipped.
  • Removed the PR-number suffix from the test name.

Backward compatibility

No API or semantic changes for users. This only disables an unsafe optimization in a specific NaN case to avoid incorrect filtering.

Tests

• Added regression test test_float_geq_nan_skip_batch_predicate.
• Local tests for the modified module pass.

Notes

• Hypothesis (the parametric test) discovered the failing example (e.g., [NaN, null, 0.0, null]). The root cause was introduced when enabling rowgroup/rowgroup skipping for float columns (feat: Enable rowgroup skipping for float columns #26805).
• Recommend merging this PR before the CSV-related PR so the CSV PR CI will be stable.

1. I used an AI assistant to analyze the predicate logic and craft the fix.
2. I confirm that I have reviewed all changes myself, and I believe they are relevant and correct.

[ritchie46]

This is already fixed in main. Your other PR can rebase on main and that should be fine.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 81.27%. Comparing base (9d8a77e) to head (ab83baa).
⚠️ Report is 5 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #27675      +/-   ##
==========================================
- Coverage   81.48%   81.27%   -0.22%     
==========================================
  Files        1840     1840              
  Lines      256109   256110       +1     
  Branches     3181     3180       -1     
==========================================
- Hits       208696   208145     -551     
- Misses      46588    47139     +551     
- Partials      825      826       +1     

☔ 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.