refactor(amber): stop hardcoding S3 in REST catalog init (#4988)

### What changes were proposed in this PR?

Stop hardcoding `s3.endpoint`, `s3.region`, `s3.path-style-access`,
`s3.access-key-id` and `s3.secret-access-key` at REST-catalog init in
both `IcebergUtil.createRestCatalog` (Scala) and
`iceberg_utils.create_rest_catalog` (Python). Both helpers now pass only
`warehouse` + catalog `uri` (and on the Scala side the `FileIO` impl
hint).

**Why:** When a Lakekeeper warehouse is created, its S3 settings
(endpoint, region, credentials, path-style) are registered against that
warehouse on the server. At catalog init the client only needs
`warehouse` + `uri` — Lakekeeper resolves the S3 config from the
warehouse record and serves it back. The hardcoded `StorageConfig.s3*`
values on the client were redundant, and forcing them everywhere also
pinned every warehouse to the single system bucket. Removing them lets
each warehouse own its own storage settings.

`StorageConfig.s3*` itself is kept —
`pytexera/storage/large_binary_manager.py` still uses it for the
non-Iceberg `texera-large-binaries` bucket (R UDF large-binary support),
which is out of scope.

### Any related issues, documentation, discussions?

Closes #4987

### How was this PR tested?

- `sbt "WorkflowCore/compile"` — passes; verifies no other Scala caller
depends on the removed properties.
- Python edits parse cleanly via `ast.parse`; the only caller
(`iceberg_catalog_instance.py`) is updated to match the new
`create_rest_catalog` signature.

End-to-end verification (warehouse with its own S3 settings → REST
catalog opened with only `warehouse` + `uri` → table round-trip)
requires a running Lakekeeper, which CI doesn't have today. #4276
(draft) wires Lakekeeper into CI; once that lands I'll add the
integration test on top of it.

### Was this PR authored or co-authored using generative AI tooling?

Generated-by: Claude Code (Opus 4.7)

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>