fix(coverage): exclude convergence tests from map build; harden refresh bot commit

[sbryngelson]

What

A manual run of the weekly coverage-refresh workflow (workflow_dispatch) surfaced two automation bugs. The SLURM gcov collector itself ran correctly in CI (618 tests → 574 mapped), but:

1. The map builder ran convergence tests and they all failed (32/618).
--build-coverage-map passed every case to the direct-invocation collector, including kind=convergence cases. Convergence tests are order-of-accuracy checks driven by convergence.py, which fills in grid resolution and patch geometry per refinement level at runtime — their base params are skeletons (m=n=p=0, no patch_icpp%geometry) that cannot run standalone. So all 32 errored in pre_process (geometry must be set / nGlobal < min_cells).

• Map impact: none harmful. Failed tests are omitted → absent from the map → select_tests conservatively always-runs them (rung 5). That is the correct behavior for convergence checks anyway. This change just makes it explicit and removes 32 spurious failure warnings.

2. The commit-back ran the pre-commit hook and aborted before pushing.
git commit triggered the repo hook (./mfc.sh precheck/spelling) against the clean: false working tree (polluted with leftover SLURM coverage artifacts), which failed → commit aborted → the git push never ran. The bot commit stages only the binary map, so it should bypass source-code prechecks: git commit --no-verify.

Changes

• toolchain/mfc/test/test.py: exclude kind=='convergence' cases from the coverage-map build; log how many are excluded (always-run by design).
• .github/workflows/coverage-refresh.yml: git commit --no-verify for the map-refresh bot commit.

Verification

• {'golden': 586, 'convergence': 32} — filter excludes exactly the 32 tests that failed in CI.
• ./mfc.sh precheck + ./mfc.sh lint (273 unit tests) pass locally; spell check passes locally, confirming the CI hook failure was a runner-tree artifact, not a master regression.

Still unverified

The CACHE_PUSH_TOKEN branch-protection bypass never executed (the commit aborted first). The next refresh run after this merges will exercise the push for the first time.

[Copilot]

Pull request overview

This PR fixes two issues in the weekly coverage-refresh automation: (1) convergence tests (which require the convergence.py driver) were incorrectly being passed to the direct-invocation coverage map builder and failing, and (2) the bot’s commit step could be aborted by repo pre-commit hooks when the runner working tree contains non-source artifacts.

Changes:

• Exclude kind == "convergence" cases from the coverage-map build path and emit a warning indicating how many were excluded.
• Bypass repo commit hooks for the bot commit by using git commit --no-verify in the coverage-refresh workflow.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
toolchain/mfc/test/test.py	Filters out convergence cases during --build-coverage-map so only directly runnable cases are mapped.
.github/workflows/coverage-refresh.yml	Uses git commit --no-verify to prevent runner-tree artifacts from aborting the automation commit.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 61.31%. Comparing base (574e53d) to head (b84862a).
⚠️ Report is 2 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1464   +/-   ##
=======================================
  Coverage   61.31%   61.31%           
=======================================
  Files          72       72           
  Lines       19771    19771           
  Branches     2852     2852           
=======================================
  Hits        12123    12123           
  Misses       5699     5699           
  Partials     1949     1949           

☔ 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.