Remove Squad workflows; cost/marker/typing/report cleanup

[sbroenne]

Second cleanup PR acting on the expert repo review.

Removed

• Squad CI workflows — deleted the 11 squad-*.yml / sync-squad-labels.yml workflows plus the .squad/templates/workflows/ copies that could regenerate them.

Added

• Cache-read tokens now priced — estimate_cost() gains a cache_read_tokens arg priced via an optional per-model cache_read rate in pricing.toml (defaults to 0.0, so no change unless configured). Wired through fixtures._convert_to_aitest.
• pyright now covers tests/ — include = ["src", "tests"] (excludes tests/visual, which needs the optional playwright extra); CI runs config-driven uv run pyright. Fixed the surfaced type errors in the scoring/plugins tests and the event-mapper helper.
• tests/unit/test_cost.py — pure-logic coverage for the cost arithmetic (incl. cache-read), unknown-model handling, zero-token short-circuit.

Changed

• --strict-markers enforced — unregistered markers now fail collection.
• Report generation resiliency — in pytest_sessionfinish, insight/HTML/MD rendering is wrapped so a failure no longer discards the already-written JSON, skips the min-pass-rate gate, or bypasses cleanup.

Verification (WSL)

• ruff check + format: clean
• pyright (src + tests): 0 errors
• unit tests: 526 passed (6 new cost tests), 8 snapshots passed
• full-suite collection with --strict-markers: 612 tests, 0 marker errors

[github-actions]

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Scanned Files

• .github/workflows/squad-insider-release.yml
• .github/workflows/squad-release.yml
• .github/workflows/sync-squad-labels.yml