Relax REST client /v1/config response constraints

[ccancellieri]

fixing #2122

Rationale for this change

Are these changes tested?

Yes tested locally

Are there any user-facing changes?

Nope, just use them

[Fokko]

Thanks @ccancellieri for raising this PR. This is something that ideally should be fixed on the BigLake side, but there is no harm to be a bit more relaxed about deserializing these messages. Thoughts @kevinjqliu ?

[kevinjqliu]

This is something that ideally should be fixed on the BigLake side, but there is no harm to be a bit more relaxed about deserializing these messages.

+1 BigLake REST catalog API should be fixed in order to be compliant with the Iceberg REST Catalog spec, which requires both defaults and overrides fields in the /v1/config response.
https://github.com/apache/iceberg/blob/b7154119f97870608429d5a9950aaaad6d2a0276/open-api/rest-catalog-open-api.yaml#L1937-L1942

That said, we've seen this issue with Nessie IRC as well (see #1524). As an IRC client, pyiceberg can be more relaxed about these fields.

[kevinjqliu]

this ones odd, i think we should revert this change. The listNamespace response should return an empty namespaces list

https://github.com/apache/iceberg/blob/b7154119f97870608429d5a9950aaaad6d2a0276/open-api/rest-catalog-open-api.yaml#L3967-L3976

[kevinjqliu]

cc @rambleraptor @talatuyarer, what do you think?

[Fokko]

That said, we've seen this issue with Nessie IRC as well (see #1524). As an IRC client, pyiceberg can be more relaxed about these fields.

This PR can be used as an interim solution, but other clients will probably error out , so eventually this has to be fixed in BigLake :)

[kevinjqliu]

Thanks @ccancellieri for the PR and @Fokko for the review :)