Add chat finetuning in custom model mappings so e2e tests don't fail

[innainu]

We want to test chat finetuning internally, but python-sdk e2e tests fail if we run a chat finetune in staging due to an unrecognized type. Chat finetunes are protected with feature flags, so users will get an error if they try to call. Can make the error more explicit in the sdk if needed.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}