fix: terminate backends before dropping databases and close pg-cache pools in codegen

[pyramation]

Summary

Fixes the systemic dropdb failure when ephemeral databases have lingering connections:

dropdb: error: database removal failed: ERROR: database "codegen_pgpm_..." is being accessed by other users
DETAIL: There are 2 other sessions using the database.

Two-layer fix:

1. Safety net in DbAdmin.safeDropDb() (pgsql-client): When dropdb fails due to active sessions, terminates backends via pg_terminate_backend and retries — only in the catch path, so the happy path is unchanged and we get log visibility into when this happens.

2. Proper pool cleanup in codegen PGPM paths (graphql/codegen): deployPgpm() and getPgPool() cache connections in pg-cache that were never released before the ephemeral database was dropped. Now explicitly calls pgCache.delete(dbName) + waitForDisposals() before teardown in both PgpmModuleSchemaSource.fetch() and generateMulti() shared source cleanup.

Review & Testing Checklist for Human

• Verify pgCache.delete() key matches getPgPool() key: getPgPool() caches by config.database (the database name). The cleanup uses dbConfig.database / shared.ephemeralDb.config.database. Confirm these always resolve to the same key.
• PostGraphile internal pool: buildSchemaSDL() → makeSchema() creates its own connection via makePgService that is not in pg-cache. The safeDropDb terminate-and-retry is the only safety net for these. Consider whether makeSchema should be patched to release its pool, or if the safety net is sufficient.
• Test with a real codegen PGPM workflow: Run a codegen that uses pgpmModulePath or pgpmWorkspacePath and verify the ephemeral database is cleanly dropped without the "is being accessed" warning.

Notes

• The safeDropDb terminate-and-retry pattern already existed in pgpm test-packages and pgpm kill commands — this just brings it to the core DbAdmin class so all consumers benefit.
• The retry is single-attempt. In theory, new connections could be established between pg_terminate_backend and the retry dropdb, but this is extremely unlikely during test/codegen cleanup.

Link to Devin session: https://app.devin.ai/sessions/ca50a2babd834e40a9d5d00b379bd459
Requested by: @pyramation

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring