Fix setup_handler KeyError and migrate owner_type to DB enum

[hiroshinishio]

Summary

• Fixed KeyError: 'sender_id' in setup_handler that caused setup to silently fail (no Slack notification) for new installations like joeHeld/Unit-Testing-Demo
• Properly populated all required BaseArgs fields (sender_id, sender_email, sender_display_name, is_fork, pr_title, pr_body, etc.) matching patterns from other handlers
• Migrated owner_type column from text to PostgreSQL enum (owner_type_enum) in both dev and prod databases across installations, owners, and usage tables
• Updated schema generator to auto-read PostgreSQL enum values from pg_enum and emit Literal types
• Added X-Sender-Id header to the /setup_coverage_workflow API endpoint

Social Media Post (GitAuto)

New user installed GitAuto on a single repo. Setup ran, cloned the repo, created a PR branch - then silently crashed. No Slack alert, no error visible to the user. The setup handler was missing a required field (sender_id) that every other handler populates. While fixing it, we found the root cause was deeper: owner_type was a text column in our DB, allowing empty strings and typos like "O" instead of "Organization". Migrated it to a PostgreSQL enum so the DB enforces valid values.

Social Media Post (Wes)

Spent an hour tracing why a new user's setup created a PR branch but never finished. CloudWatch showed a KeyError buried inside an exception handler that swallowed it silently. The fix was straightforward - pass sender_id through to the handler. But digging into the data, I found 227 rows in dev and 14 in prod with empty or malformed owner_type values. The column was plain text with a default of empty string. Changed it to a Postgres enum. No more silent garbage.