[iceberg-rest-fixture] fix sqlite error under high concurrent load

[kevinjqliu]

Closes #13366

Use file-based sqlite db (i.e. jdbc:sqlite:/tmp/iceberg_catalog.db) instead of in-memory sqlite db (jdbc:sqlite:memory:)
File-based sqlite db allows for concurrent access with multiple jdbc connections. By contrast, in-memory sqlite db creates a new db per jdbc connection. Even with shared caching enabled (jdbc:sqlite:file::memory:?cache=shared), in-memory sqlite db can perform concurrent reads but concurrent writes will have race conditions and fail a subset of the requests.

Test

Tested locally by explicitly overriding the env parameters in pyiceberg's integration test setup

      - CATALOG_URI=jdbc:sqlite:/tmp/iceberg_catalog.db

and then running the "Reproduce" code in #13366

[Fokko]

Thanks @kevinjqliu , this change makes sense to me.

Next to the concurrency issues, I think it also makes sense to keep the data around when restarting a container.