Add pipeline_runs table population — table exists in schema but is never written to

[haydenk]

Problem

Migration 0003_create_pipeline_runs.sql creates a pipeline_runs table, but no code in any crate inserts rows into it. The table is completely unused. Either the feature was planned but never implemented, or it was removed from the application code without removing the migration.

Expected behavior

Either:

• Implement pipeline_runs inserts: the downloader and ETL should write a row at the start of each run and update it on completion with duration, row counts, and status.
• Or: remove migration 0003 and the table entirely if the design has moved to tracking state purely via esiid_documents.