ci: pin uv, disable fail-fast, add job timeouts, drop dead conditions

[GitBib]

Follow-up to the setup-uv v9 bump in #121 plus a few things found while clearing the dependabot queue.

prune-cache: true — v9.0.0 flipped the prune-cache default from true to false, which is the release's only breaking change. enable-cache: true is set in all four workflows, so the change applies here: the uv cache stops being pruned and grows. With a 15-job matrix against a 10 GB per-repo cache limit and LRU eviction, that turns into cache thrash rather than a hard failure. This restores the old behaviour explicitly.

fail-fast: false — one failing job used to cancel the other 14. On #126 macOS 3.12 failed and everything else was cancelled, so there was no way to tell from the run whether the problem was platform-specific without digging into logs. Both matrices now report independently.

Pinned uv to 0.12.1 — version: latest meant a uv release could break CI with no change in the repo, and the failure would look like ours. The four call sites now agree on a version.

timeout-minutes on every job — there was none anywhere, so the default of 360 applied. A hung choco install would burn six hours of runner time.

Dropped dead if: clauses:

• python-tests.yml gated codecov on workflow_dispatch, which Tests never receives — it triggers on push to master and pull_request only.
• python-docs.yml gated deploy on pull_request and the job on !github.event.pull_request.draft, but Docs only triggers on push to master and workflow_dispatch. Reading it, you would think docs build on PRs. They do not.

Verified prune-cache exists as an input in setup-uv v9.0.0 (default: "false") rather than assuming it from the changelog. All four workflow files parse, and every job now carries a timeout.

Docs publish only from master — the deploy step was gated on the event name, and workflow_dispatch can be started on any branch, so a manual Docs run from a feature branch would have published that branch's build to gh-pages. Now gated on github.ref == 'refs/heads/master', which covers both push-to-master and a manual run on master. The build still runs on dispatch from any branch, it just does not publish.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.11%. Comparing base (46a123b) to head (4cf631b).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #128   +/-   ##
=======================================
  Coverage   99.11%   99.11%           
=======================================
  Files          21       21           
  Lines        1477     1477           
=======================================
  Hits         1464     1464           
  Misses         13       13           

☔ 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.