[Backend] Generate and upload changelog JSON when a new dataset version is detected

[cka-y]

Describe the problem

There is no persistent artifact capturing what changed between two subsequent datasets for a given feed, and no automated process to generate and store it.

Proposed solution

Create a new Cloud Function gtfs-change-tracker (in functions-python/gtfs_change_tracker/) that orchestrates change tracking for MobilityData feeds. It is HTTP-triggered and handles orchestration via Cloud Tasks.

Responsibilities:

1. Receive feed_id, previous_dataset_id, current_dataset_id
2. Resolve GCS URLs for both datasets
3. Call the gtfs_diff module with those URLs
4. Upload the resulting changelog JSON to GCS at:
   <dataset_bucket>/<feed_stable_id>/changelogs/<previous_dataset_id>_<current_dataset_id>_changelog.json
5. Write one row to gtfs_dataset_changelog with the GCS URL

This function is intentionally MobilityData-specific. All generic diffing logic lives in gtfs_diff (see Issue 2a).

Alternatives considered

• Inline the diff logic here: rejected in favour of keeping gtfs_diff reusable and portable.