v0.6.38: Explain consensus recommendations

[MVtag]

Adds an accessible, on-demand explanation to each v0.6.37 Consensus Recommendation badge.

Users can activate a recommendation badge with mouse, touch, Enter, or Space to reveal why the filter is Recommended, Seat tune, Review, or Guarded. The inline explanation shows the combined recommendation reason, current Confidence result/score, and the concrete Position Robustness evidence. The expanded state is reflected through aria-expanded, and the summary plus Multi-Position guidance now explain that badges are interactive.

The runtime-confirmed v0.6.37 recommendation matrix is unchanged. For the current Passenger/Raw setup it remains 3 Recommended + 1 Guarded, with Confidence 97/97/100 and 3 Robust + 1 Guarded.

This is frontend-only and advisory. It does not change feature detection, Confidence, Robustness, recommendation classification, Guards, filter On/Off, Prediction, Apply Preview, Apply/Restore, snapshots, channel EQ, Store, backend APIs, or Goldhorn control. The active panel entrypoint remains kcc-soundlab-panel-0631.js; no wrapper or backend module is added.

Validation completed locally:

• Python syntax
• JSON parsing
• all frontend JavaScript syntax
• frontend cache-key consistency
• unchanged recommendation matrix
• mouse/touch/keyboard explanation contract
• clean diff check

Version bumped to 0.6.38.