Add job_id to table schemata.

[whilo]

This fixes glitches where results in one view are from a different job_id than results in another view.

Root Cause

1. Topic Storage: Topic names are stored in DynamoDB with simple keys like layer0_5 without job ID versioning
2. Non-Selective Retrieval: The system fetches ALL topic names for a conversation regardless of which job generated them
3. No Job Correlation: There's no mechanism to ensure topic names and cluster data come from the same UMAP run