Enhancement: Diagnostics verdict feedback loop for adaptive debounce

[Unthred]

Problem

Adaptive debounce (#7) learns from save bursts; diagnostics verdicts could refine learning (noted out of scope in #7).

Acceptance criteria

• Verdict (helpful / noisy rebuild) stored per trigger journal entry
• Optional influence on debounce learning weights
• Unit tests for feedback calculator

Surfaces

Infrastructure, TrayApp, docs

Related

#7 (Done), #10 (Done)