feat(sessionize): quantify sessionization debug reasons and cover fli…

[Sarcastic-Soul]

What this changes

Closes #18.

Summary

Extends the sessionization debug info (_debug block emitted when --debug / debug=True is enabled) with quantified reason strings and coverage for deterministic boundaries (§5):

1. Quantified Reason Strings:
   • session_gap: Formatted with measured gap against §5.2 threshold (session_gap: 3600s > 300s).
2. Deterministic Boundary Coverage:
   • merged flicker: Notes folded interruptions and their span vs §5.3 limit (merged flicker: Slack 8s <= 20s).
   • dwell capped: Notes when frame active time is capped at §5.1 limit (dwell capped: 90s).
   • device: Notes when frames are partitioned by monitor device (device: Display 1).

Remains Tier-1: measured, reproducible, namespaced under _debug, opt-in via --debug, and byte-identical when off.

Checklist

• Output stays deterministic (no clock, network, randomness, or model calls)
• New behavior has a test; pytest -q passes
• New schema fields, if any, are measured - or live in the tier-2 inferred block with a confidence tag
• No typed text or page content is emitted by default

[nossa-y]

Debug-off output is byte-identical - good. But the flicker note quotes the wrong quantity: the SPEC 5.3 gate is B's span (8s on the fixture), the note prints interruption seconds - so it emits "Slack 20s <= 20s", and an A-B-A with a dwell-capped return prints "B 100s <= 20s", which is false. Your PR body's own example ("8s <= 20s") is the correct output.

1. Compare wall_seconds() in the note; update tests to cover span != active.
2. Add a device-boundary test (SPEC 5.5, see test_multidevice.py).
3. Only emit the device note when the window spans >1 device - as-is every entry on a single-monitor DB gets "; device: monitor_1".
4. Drop Segment.device - written, never read.