[Good First Issue] Validate start_date <= end_date in PredictRequest

[femi23]

Background

PredictRequest (in src/climatevision/api/schemas.py / api/main.py) accepts a start_date and end_date, but currently does not check that the range is well-ordered. A request with start_date="2026-06-01" and end_date="2026-01-01" is silently accepted and produces an empty image stack downstream.

Acceptance criteria

• Add a Pydantic v2 model_validator(mode="after") (or field_validator) on PredictRequest that raises ValueError if start_date > end_date.
• Update the OpenAPI description for both fields to make the ordering explicit.
• Add a pytest in tests/ that:
  • posts a valid range and asserts the request reaches the inference layer (mock the inference call),
  • posts a reversed range and asserts the API returns 422 Unprocessable Entity with the validator's error message in the response body.

Pointers

• The schema lives in src/climatevision/api/schemas.py. The route handler is @app.post("/api/predict") in api/main.py.
• For the test, use fastapi.testclient.TestClient and patch run_inference_from_gee so the test does not hit GEE.
• Existing validators (e.g. on bbox) are good reference patterns.

Why this is a good first issue

Self-contained: one validator + one test. No need to understand the inference pipeline, model loading, or GEE. Done in well under 50 lines.