fix: Add error handling to write_predictions_to_csv

[exploring-curiosity]

Closes #17

What changed

The function writes predictions to CSV without any validation or error handling. If predictions is None, empty, or malformed, it will crash or produce a broken file. Add input validation (check predictions is a non-empty list of tuples) and wrap file I/O in try/except to handle permission/disk errors gracefully.

Approach

Locate the write_predictions_to_csv function in the codebase (likely in a Python file under model_serving or data_pipeline). Add input validation to ensure predictions is a non-empty list of tuples, raising a ValueError if invalid. Wrap the CSV writing logic in a try/except block to catch and handle IOError/OSError (e.g., permission denied, disk full) with a descriptive error message.
Files: model_serving/BirdClassification/app.py

Verification

Tier	Result	Detail
T0 Diff sanity	pass	internal: diff sanity
T1 Static	skipped	no notebooks changed
T2 Smoke	skipped	no build step configured
T3 Tests	skipped	pre-existing failure, not introduced by this change
T4 Runtime	skipped	no runtime check configured

Why this is a draft

nothing beyond diff sanity was verified - refusing to merge unverified work

Note: T1 (no notebooks changed), T2 (no build step configured), T3 (pre-existing failure, not introduced by this change), T4 (no runtime check configured) - these were not verified.