Release EvidenceForge 1.13.0

[DavidJBianco]

Summary

Release EvidenceForge 1.13.0 with canonical lifecycle, identity, observation,
remote-authentication, and cryptographic planning contracts, plus long-scenario
performance and reliability improvements.

What changed

• Centralized canonical network transaction, sensor observation, explicit proxy,
  identity lifecycle, session closure, Linux sudo, and Windows remote-authentication
  planning.
• Added canonical cryptographic material, standards-valid OCSP transactions,
  valid DKIM signing, and rendered cryptographic evidence contracts.
• Replaced global connection scans with indexed tuple lookup and close-time
  eviction.
• Replaced emitter barrier polling with FIFO acknowledgements; the supplied
  48-hour workload improved by 7% with byte-identical output.
• Unified active/ended sessions, processes, threads, connections, expiry caches,
  Linux PID history, and logind history behind reusable duration-stable indexes.
• Replaced the quadratic Linux PID collision walk with bounded deterministic
  allocation.
• Prevented backdated bootstrap from resurrecting an ended session, fixing the
  supplied 56-day scenario's late lifecycle failure.
• Bounded syslog retention with record-preserving hourly spools.
• Replaced the duration-growing sensor-UID correlation table with a constant-size
  completed-connection handoff, and compacted stale grouped temporal-index records.
• Stabilized Rich ETA estimation over a 15-minute window.
• Moved COLLECTION_PROFILE.json from the ingestible data/ tree to the
  run-metadata root, including transactional overwrite and rollback handling.
• Included the Dependabot updates for actions/setup-python 7, pre-commit 4.6.1,
  Ruff 0.15.22, and Typer 0.27.0.

Determinism and compatibility

Repeated runs using the same version remain deterministic. The bounded Linux PID
repair intentionally changes PID-derived output relative to earlier development
builds.

For the final network-retention change, the supplied scenario was generated for
48 hours from exact parent commit a6e79b8c and from the new implementation.
All 24 artifacts and the complete directory trees were byte-identical. Moving
COLLECTION_PROFILE.json is an intentional path-only packaging change; its JSON
content is unchanged.

Validation

• Default gate: 4,988 passed, 41 skipped in 240.51s.
• uv run pytest --no-cov --include-slow: 5,001 passed, 28 skipped in 300.94s.
• Final focused CLI/engine/index/dispatcher suite: 170 passed.
• High-volume regression: 100,000 network-identifier publications retain only
  the latest connection's two format values.
• Real CLI smoke generation verified COLLECTION_PROFILE.json at the output root
  and absent from data/.
• uv run ruff check .
• uv run ruff format --check .
• git diff --check

Coverage was not rerun locally at maintainer direction because coverage
instrumentation makes generation-heavy tests prohibitively slow. CI retains the
coverage gate; slow tests are run without coverage per project policy.

Release behavior

After merge, the release workflow should verify version consistency and tag
availability, create annotated tag v1.13.0, and publish the GitHub Release.