fix(ml-service): return 422 for non-numeric/NaN fields in /predict-yield

[saidai-bhuvanesh]

What changed

In ml-service/app.py, the POST /predict-yield handler converted request fields with bare float(...) calls and no error handling:

• Non-numeric values (e.g. {"area_hectares": "abc"}) raised ValueError → generic 500.
• NaN/Infinity values parsed fine but propagated through temp_modifier/rain_modifier into round(expected_yield, 2), yielding invalid JSON.

Fix

Added parse_yield_numeric(), which validates each present numeric field (area_hectares, avg_temperature, expected_rainfall):

• non-numeric → 422 with {"error": "Validation failed", "details": [...]}
• NaN/inf → 422
• missing fields keep the existing default behaviour (no regression)

This matches the structured 422 pattern already used by the /predict and /quality handlers.

Tests

Added ml-service/test_predict_yield.py (unittest) covering:

• non-numeric area_hectares → 422
• NaN area → 422
• non-numeric avg_temperature → 422
• Infinity rainfall → 422
• valid request → 200
• missing numeric fields (defaults) → 200

All 6 tests pass (ML_API_KEY=test-key python -m unittest test_predict_yield).

Closes #1234

---

This PR was created by an AI agent (OpenHands) on behalf of @saidai-bhuvanesh.

[vercel]

@openhands-agent is attempting to deploy a commit to the Nitya Gosain's projects Team on Vercel.

A member of the Team first needs to authorize it.