Add .codecov.yml with threshold-based coverage status checks

[khatchad]

Summary

Adds a minimal .codecov.yml so Codecov's project/patch status checks tolerate small base-vs-head coverage drift instead of failing on every fractional drop. Mirrors the configuration in ponder-lab/ML.

Why

Without an explicit .codecov.yml, Codecov uses default target: auto with zero tolerance: any project-coverage decrease at all triggers codecov/project: FAILURE. Base-vs-head test-environment drift or flakiness can shave 1-2 lines off the project total and fail the gate even on PRs that don't touch Java code.

Concrete case: #516 (the POM cleanup mirror of ponder-lab/ML#272) has codecov/patch: SUCCESS (no Java changed, so the patch is trivially 100% covered) but codecov/project: FAILURE from a -0.03% project delta (-2 hits / +2 misses) that can't be attributed to the diff. The default Codecov policy fails on any non-zero negative delta; this configuration absorbs the noise.

What This Changes

coverage:
  status:
    project:
      default:
        target: auto
        threshold: 1%
        informational: false
    patch:
      default:
        target: auto
        threshold: 2%
        informational: false
github_checks:
  annotations: true

• project threshold 1%—tolerates project-coverage drops of up to 1% before failing. Real regressions still fail.
• patch threshold 2%—same idea for changed-line coverage.
• github_checks.annotations: true—surfaces uncovered changed lines as inline annotations in the PR's "Files changed" view.

informational: false keeps the checks as real build signal rather than unconditionally green—the thresholds are the absorptive layer. The per-PR Codecov comment remains enabled by default (Codecov's default behavior with comment unspecified).

History Of The Equivalent ponder-lab/ML Config

For reference, the four commits that produced this state on the fork:

• f7f610ff (#243): added .codecov.yml with informational: true (unconditionally green).
• fd19bb91 (#263): disabled per-PR comment, enabled inline annotations.
• ee247f0b (#264): flipped informational: false with the 1%/2% thresholds—"Now that there's no comment, we need to see the build failure."
• 3eac854a (#266): re-enabled the per-PR comment.

This PR ports the current combined state in one commit.

Test Plan

• After this lands, retry Drop dead logging.properties include from jython3.test POM (#493 follow-up) #516 (or any past PR with a noisy codecov/project: FAILURE)—the threshold should absorb the drift and the gate should pass.
• No effect on actual coverage measurement; only the pass/fail decision on the GitHub status check changes.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Adds a repository-level Codecov configuration to make codecov/project and codecov/patch status checks tolerant to small coverage fluctuations, reducing CI noise while keeping coverage regressions as real failures.

Changes:

• Add .codecov.yml configuring project and patch coverage statuses to use target: auto with threshold tolerances (1% / 2%).
• Enable github_checks.annotations so uncovered changed lines appear as inline PR annotations.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 57.90%. Comparing base (ad96543) to head (ab803d8).

Additional details and impacted files

@@             Coverage Diff              @@
##             master     #529      +/-   ##
============================================
- Coverage     57.93%   57.90%   -0.03%     
+ Complexity      625      624       -1     
============================================
  Files           111      111              
  Lines          7671     7671              
  Branches        856      856              
============================================
- Hits           4444     4442       -2     
- Misses         3049     3051       +2     
  Partials        178      178              

☔ 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.