Allow r2:// URL scheme

[timmclaughlin]

Description

Allow Cloudflare R2 (`r2://`) URLs through pg_lake's PG-side URL allowlists.

Why this works

The DuckDB `S3FileSystem` natively recognizes `r2://` alongside `s3://`, `s3a://`, `s3n://`, `gcs://`, and `gs://` — see `duckdb-httpfs/src/s3fs.cpp` `S3FileSystem::CanHandleFile`. `duckdb_pglake`'s `RegionAwareS3FileSystem` already short-circuits all non-`s3://` URLs straight to the underlying `S3FileSystem` (region auto-detection is intentionally `s3://` only — R2 always uses `auto`), so no `duckdb_pglake` change is needed. The only thing gating `r2://` today is pg_lake's PG-side URL allowlist.

What this PR changes

1. Add `r2://` to both URL allowlists:

   • `pg_lake_engine/src/copy/copy_format.c` `IsSupportedURL` — the project-wide check consulted by:
     • `pg_lake_copy` (`CREATE TABLE ... load_from`/`definition_from`)
     • `pg_lake_iceberg` (`lake_iceberg.metadata`/`files`/`snapshots`, the `pg_lake_iceberg.default_location_prefix` GUC)
     • `pg_lake_table` (`file_size`/`file_exists`/`file_preview`/`delete_file`, foreign-table `path`/`location` options, `CREATE TABLE` for iceberg / pg_lake_table)
     • `pg_lake_benchmark`
   • `pg_lake_table/src/util/s3_file_utils.c` `IsFileListSupportedUrl` — the tighter list-only check used by `lake_file.list` (excludes `http(s)://` since those can't enumerate directories).

2. Refresh the now-stale `"only s3://, gs://, az://, azure://, and abfss://"` error strings across the codebase so they accurately reflect what `IsSupportedURL` accepts. This also picks up the previously-missing `hf://` — HuggingFace URLs have been accepted by `IsSupportedURL` for a while but the error strings were never updated. Updated the four pytest assertions that match those strings exactly.

Repro of the original failure

```sql
SELECT * FROM lake_file.list('r2://my-bucket/prefix/*');
-- ERROR: list_files: only s3://, gs://, az://, azure://, abfss://, and hf:// urls are supported
```

After this PR the same call falls through to `pgduck_server` / DuckDB which handles `r2://` via the standard R2 secret mechanism.

---

Checklist

• I have tested my changes and added tests if necessary
• I updated documentation if needed
• I confirm that all my commits are signed off (DCO)

---

Notes for reviewers

• I considered narrowing the patch to only `IsFileListSupportedUrl` + the `list_files` error string, but every other call site that gates on `IsSupportedURL` (CREATE FOREIGN TABLE, iceberg metadata functions, the location-prefix GUC, etc.) would have inconsistent behavior — `r2://` would work in some paths and not others. Adding `r2://` centrally in `IsSupportedURL` keeps the surface consistent.
• The error-message refresh is mechanical: every occurrence of the stale prefix list now lists what `IsSupportedURL` actually accepts. Happy to split that out of this PR if you'd prefer two smaller commits.
• Did not add an e2e test that hits a real R2 bucket — that would require account credentials in CI. The pytest assertions in `test_list_file.py`, `test_file_preview.py`, `test_create_table.py`, `test_create_as_select.py`, `test_create_iceberg_table_load_from.py`, `test_iceberg_functions.py`, `test_queries.py`, and `test_writable_iceberg.py` exercise the allowlist gate against invalid URLs and now verify the updated error string. Happy to add an R2-specific positive test if you have a preferred pattern.