C10: Coherence wired with real SelfFingerprint

[filipeforattini]

Parent

#46

What to build

Wire Coherence::compute_coherence against a real SelfFingerprint populated from live capture (C8) and compared against the real catalog (C9). After this slice, every FingerprintReport produced by the engine carries a populated Coherence block with our_ja3_matches_profile, their_antibot_compatible_with_our_profile, and warnings reflecting actual measurements rather than placeholder values.

Fingerprinter::analyze_hot and analyze_warm both accept an optional SelfFingerprint argument (or read from a Fingerprinter-held field set via builder) so the coherence cross-check runs whenever a fingerprint is available. The constructed scenarios from B13 keep passing; new tests cover the integration with real catalog data.

Acceptance criteria

• Fingerprinter holds an optional Arc<SelfFingerprint> set via with_self_fingerprint(fp)
• Fingerprinter::analyze_hot populates report.coherence when self-fingerprint is set
• Coherence.our_ja3_matches_profile reports Some(true) on a clean run with measured catalog hashes (C9)
• Coherence.our_ja3_matches_profile reports Some(false) + drift warning when the live capture diverges from catalog
• Coherence.their_antibot_compatible_with_our_profile populates from the antibot Detections in the same report
• All existing coherence::tests pass
• New integration test exercises the full pipeline: live capture → catalog diff → coherence in report
• NDJSON regression byte-stable

Blocked by

• C8: ClientHello byte capture + SelfFingerprint::capture_live #50
• C9: Real catalog hashes from live capture #54