fix(migrations) make resumable and safe against schema drift

From the PR description

Summary

Makes the production SQL migration process resumable and safe against schema drift in the existing Supabase database.

Why / Motivation

Production already contained parts of the document-version schema, while several historical migrations assumed older documents columns still existed. Re-running the migration sequence therefore failed on references such as d.filename, d.storage_path, and changed RPC return types.

Changes

• Added a migration ledger (public.mike_schema_migrations) so each successful migration is recorded and skipped on subsequent deploys.
• Updated the migration runner to execute only unapplied migration files and reload the PostgREST schema after completion.
• Guarded document-version backfills that depend on legacy documents metadata columns (filename, storage paths, file type, size, page count).
• Made affected migrations safe when columns or tables have already been introduced.
• Updated overview RPC migrations to drop incompatible existing function signatures before recreating them when the returned row shape changes.
• Improved migration execution so a failed deploy resumes at the failed migration rather than rerunning the entire history.

Tradeoffs & risks

• The migration ledger treats a migration as complete only after its SQL file succeeds. A failed migration must be fixed and rerun.
• Dropping and recreating overview RPC functions creates a very short availability gap for those RPC endpoints during deployment.
• Guarded legacy backfills intentionally skip data-copy work when the old source columns no longer exist; this prevents deployment failure but does not recreate removed legacy data.
• This does not replace a full schema-diff or formal migration framework; it hardens the current ordered SQL migration workflow.

How verified

• Ran migrations against the production Supabase connection through the deployment workflow.
• Confirmed previously applied migrations are recorded and skipped on rerun.
• Confirmed migration failures now identify the exact SQL file and statement, allowing fixes to resume from that point.
• Queried pg_stat_activity through psql to confirm the database connection works and no blocking migration lock was present.

Checklist

• Ran the relevant build/test command for the area changed.
• Reviewed git diff and removed unrelated changes.
• Updated docs / env examples if setup, config, or behavior changed.
• No secrets, API keys, real documents, or .env files committed.