fix: normalize PostgreSQL URI schemes case-insensitively

## Bug description

`application/core/db_uri.py` recognizes PostgreSQL URI prefixes with case-sensitive `startswith()` checks. URI schemes are case-insensitive, so a valid operator-supplied value such as `POSTGRESQL://user:pass@example.com/db` is returned unchanged. SQLAlchemy then fails with `NoSuchModuleError: Can't load plugin: sqlalchemy.dialects:POSTGRESQL`. Likewise, `POSTGRESQL+PSYCOPG://...` is not converted to the libpq-compatible form for `PGVECTOR_CONNECTION_STRING`.

## Expected behavior

Recognized PostgreSQL schemes should be normalized regardless of their casing, while preserving the remainder of the URI and leaving unknown schemes unchanged.

## Reproduction

```python
normalize_postgres_uri("POSTGRESQL://user:pass@example.com/db")
# currently returns POSTGRESQL://user:pass@example.com/db
# expected postgresql+psycopg://user:pass@example.com/db

normalize_pgvector_connection_string("POSTGRESQL+PSYCOPG://user:pass@example.com/db")
# currently returns the unsupported SQLAlchemy dialect URL
# expected postgresql://user:pass@example.com/db
```

## Proposed fix

Match only the recognized scheme prefix case-insensitively, emit the canonical lowercase target prefix, and add regression coverage for both normalizers.

Start in application/core/db_uri.py with normalize_postgres_uri and normalize_pgvector_connection_string. Add regression coverage for the two mixed-case PostgreSQL examples, preserving the URI remainder and leaving unknown schemes unchanged; done means both normalizers emit the expected canonical prefixes.

Written by the indexing model from the issue text.