[Spec 03] Speaking-rate per segment (beta thresholds)

[fwright21]

Summary

Implement per-segment speaking-rate checks (Spec 03) with timestamps and severity for suite/UI triage.

Scope

• Compute rate per aligned segment and emit segment-level severity.
• Apply stricter logic on critical segments (red_flag, meds).
• Emit flagged_regions for UI jump-to-region and review cards.
• Ensure scoring/reporting wiring uses this output consistently.

Important Note

Current speed thresholds (too fast / too slow) are intentionally conservative and should be treated as beta. They will need calibration updates by voice style, language, and domain context before being considered stable policy gates.

Acceptance Criteria

• Segments include timestamps + rate values + severity.
• Critical segments use tighter guardrails than non-critical segments.
• UI and report suggestions reflect speaking-rate findings.
• Threshold values/config are easy to tune.