fix: accept documented LiteSpeed cache-hit headers

[Ymax27]

Summary

• cache_indicators() records X-LiteSpeed-Cache, X-LSADC-Cache, and X-QC-Cache with their exact values.
• shared_cache_hit_markers() accepts only an exact hit token for those headers. miss, bypass, mixed values, and substrings such as hitter are not hit evidence.
• A four-stage cache state machine can satisfy the cache-hit checks from X-LiteSpeed-Cache: hit. The same sequence with miss does not.
• Existing Age, X-Cache, CF-Cache-Status, Cache-Status, Akamai, and Server-Timing handling is unchanged. The header does not by itself make a response look cacheable.

Scope

Fixes #40.

LiteSpeed documents hit as served from LSCache and miss as not served from cache: https://docs.litespeedtech.com/lscache/start/

The second commit on this branch removes an accidental out-of-band promotion edit so the final diff only changes cache-hit recognition.

Validation

Commands run on this branch:

PYTHONPATH=src python3 -m pytest -q tests/test_litespeed_cache.py

Observed result: 2 passed.

ruff and mypy are not installed in this environment, so those gates were not run. The detector/evidence coverage gate was also not run here.

Checklist

• CLI changes are safe by default and documented.
• New detection behavior includes tests.
• False-positive filtering remains conservative by default.
• Observations/probes are preserved even when findings are filtered.
• Evidence schema and proof-gate mutation tests pass.
• No destructive payloads or denial-of-service behavior were added.
• I reviewed the final diff and can explain the changes I am submitting.
• Validation results in this PR are from commands actually run on this branch; no results or evidence were fabricated.
• Any assumptions I could not verify are stated explicitly instead of being presented as facts.

Made with Cursor

[tayfuryldz]

I checked this against current main and ran the full maintainer gates locally: pytest, Ruff, and mypy are clean. The diff stays within the issue scope and I did not find a blocking regression.