clientauth: nonexistent pgtle.clientauth_db_name causes indefinite login hang

## Description

If `pgtle.clientauth_db_name` names a database that does not exist, every login hangs forever. The clientauth background workers FATAL on startup, the postmaster respawns them once per second, and `clientauth_hook` waits on a shared-memory rendezvous no live worker ever completes. The GUC is `PGC_POSTMASTER`, so no one can log in to fix it.

## Steps to reproduce

1. Fresh cluster with `pg_tle` in `shared_preload_libraries`.
1. Set `pgtle.enable_clientauth = 'on'` and `pgtle.clientauth_db_name = 'ghost_db_does_not_exist'`.
1. Restart, then run `psql -c 'SELECT 1'`.

## Expected outcome

Connection succeeds (matching the existing "pg_tle not installed" fallback under `on`), or is cleanly rejected under `require`.

## Actual outcome

`psql` hangs indefinitely. Killed by `timeout 10` with exit 124 and no output.

## Analysis

Have each worker bump a `workers_ready` counter after `BackgroundWorkerInitializeConnection()` succeeds. `clientauth_hook` waits briefly on it before enqueuing, and on timeout falls back to `can_allow_without_executing` / `can_reject_without_executing`.

### Logs

```
FATAL: database "ghost_db_does_not_exist" does not exist
LOG: background worker "pg_tle_clientauth worker" (PID ...) exited with exit code 1
[repeats every second, forever]
```

Reproduce the hang with the listed cluster settings and `timeout 10 psql -c 'SELECT 1'`. Trace `clientauth_hook`, the clientauth workers, `BackgroundWorkerInitializeConnection()`, and the `can_allow_without_executing`/`can_reject_without_executing` fallbacks; done means the nonexistent database no longer causes an indefinite login hang and `on` and `require` follow their stated outcomes.

Written by the indexing model from the issue text.