chore: coverage measurement in CI + README badges, test-count reconcile, all extra

[Therealdk8890]

A low-risk packaging/docs polish pass — no library source changes.

Changes

• Coverage is now measured on every PR (non-gating). The test job runs pytest --cov=dprovenancekit --cov-report=term-missing:skip-covered on each matrix leg, and pytest-cov is added to the dev extra. There is deliberately no --cov-fail-under, so coverage is visible in the log but never blocks a merge.
• README badges: added Python versions and License badges (both from PyPI metadata), rounding out the badge row.
• Test-count reconcile: README's 380+ and CONTRIBUTING's ~430 now both read ~440, keeping the honest "a default run skips uninstalled adapters; CI's full matrix runs them all" nuance.
• all extra: a convenience aggregate that installs every framework adapter at once. Documented inline that it requires Python ≥ 3.10 (openai-agents / llama-index / crewai declare that floor) and is heavy (crewai pulls chromadb/onnxruntime); 3.9 users pick specific extras.

Verification (local, Python 3.13)

• 514 passed, 2 skipped (crewai / openai-agents not installed → correctly skipped).
• Total line coverage: 63% — this is the number the roadmap's "improve if below 90%" asks for. Integration adapters are well covered (86–96%); the thin areas are CLI/benchmark/UI-viewmodel plumbing. Raising this toward a target is a separate, larger effort (see PR discussion).
• pip install -e ".[dev]" resolves with the new pytest-cov; [all] parses and resolves.

Not done (deliberately)

• No otel extra: OTLP ingestion (otel_ingest.py) is pure standard library, so an opentelemetry-* dependency would have no consumer. The 8 existing framework adapters already each ship an extra.