Inference Profiles aren't supported

Problem/Motivation

AWS has started forcing using inference profiles for certain models in the system, like Anthropic. This renders them unusable for this provider because the model IDs can't be called directly like normal on-demand models and they're not provisioned models.

Proposed resolution

The model's inference support is returned in the initial model call, so the easiest solution would be to look for that and adjust the model IDs to use the inference profile names which can be either 'us' or 'global' depending on where the user is calling the models from.