Add CI workflow to run pytest on PRs

[bradleesand]

Summary

• Adds .github/workflows/tests.yml running the full pytest suite on Python 3.11
• Triggers on every PR, on pushes to main, and via workflow_dispatch
• Installs runtime deps from requirements.txt plus pytest, pytest-asyncio, httpx (needed by FastAPI TestClient)

Why

No general CI workflow runs the suite today — only music-search-benchmark and release workflows. Running pytest on every PR catches regressions and gates the release branch.

Gating

Requires all tests to pass — no continue-on-error, so any failure fails the check. Fixed one pre-existing broken test (tests/test_provenance_version.py, expected 0.9.20 → 1.0.0) to meet this requirement.

Test plan

• Full suite passes on this branch (605 passed, 1 skipped)
• Failures fail the check (no continue-on-error)
• Workflow appears on the PR and runs to completion
• Subsequent PRs trigger the workflow automatically

[loganbuilt]

@bradleesand
Thanks for putting this together — we do want CI coverage, but I need a couple changes before we can merge this into the release path:

1. Please change the PR base branch from main to arr_integration.

2. Please update the workflow behavior so test results are actionable for the branch:

• Right now continue-on-error: true on the pytest step allows green checks even when tests fail.
• For release-branch gating, we should either:
  • remove continue-on-error so failures fail the check, or
  • split this into a clearly non-blocking/report-only workflow with explicit naming and a separate blocking workflow for required checks.

3. Please align workflow triggers with our release branch usage:

• add arr_integration to push/PR trigger scope (or make this workflow explicitly PR-only if that’s the intent).

After those changes and retargeting to arr_integration, I’ll re-review quickly.

[bradleesand]

Thanks for the detailed review!

Rebased onto current main and addressed your points:

1. Base branch — kept as main. arr_integration was merged into main for the 1.0.0 release, so main is the active release branch. Happy to change the target if you'd prefer otherwise.
2. No masked failures — I originally added continue-on-error because there were so many failing tests at the time; I'd scoped this PR to just the CI infrastructure and planned to fix the tests in a separate PR. But since the v1.0.0 release most tests now pass on main, so I just fixed the single remaining failing test and removed continue-on-error — failures now fail the check. (That one failure was a stale assertion: test_provenance_version.py still expected 0.9.20 while pyproject is now 1.0.0.) Full run is now 605 passed, 1 skipped, 0 failed.
3. Triggers — runs on all PRs + pushes to main + workflow_dispatch, aligned with main as the release branch.

[loganbuilt]

awesome thanks!