Analysis Integration

[dangrabo]

Summary

Adds Groq-powered LLM feedback to the interview flow, with a user-selectable timing mode for when feedback is generated.

• POST /feedback/generate-session — single Groq call for full-session feedback (overall summary + per-question reviews)
• Feedback timing toggle — after each question (1 call per question) or at end of interview (1 combined call)
• Interview UX — optional View Feedback scorecard, End Interview anytime, session review panel for end-of-interview mode
• Uses existing POST /feedback/generate with gpt-oss-20b; sends transcript + per-segment A/D/V scores

Test plan

• Start interview with After each question → answer → View Feedback shows scorecard
• Start with At end of interview → answer multiple questions → End Interview → session feedback loads
• Confirm pnpm lint and pre-push hooks pass