[M1][Epic 1.3] Build FastAPI ML service with sentiment analysis endpoint

[POWDER-RANGER]

Problem

Implement production FastAPI application with sentiment analysis capability. Start with lightweight models for speed.

Tasks

• Create FastAPI app in ml/main.py or ml/src/app.py
• Implement GET /health endpoint
• Implement POST /analyze/sentiment accepting {text: string}
• Return {score: float(-1..1), confidence: float(0..1)}
• Choose sentiment model (recommend transformers pipeline or TextBlob for MVP)
• Store analysis results in Postgres via backend API call
• Add latency measurement (target <1s locally)
• Write unit tests for sentiment scoring

Acceptance Criteria

• ✅ FastAPI responds to health checks
• ✅ Sentiment endpoint returns scores within 1 second locally
• ✅ Results visible through backend /api/analytics/overview
• ✅ Confidence scores correlate with text certainty

Priority: P0 - Blocker
Labels: ml, fastapi, M1, P0