feat: Error message consistency

[iamemilio]

What does this PR do

1. Provides a set of tests to validate responses and conversations API error responses
2. Enable Capturing error recordings and replaying them as their original type in the record/replay system
3. Compares and converts Error responses to conform to OpenAI standards whenever possible
4. Introduces a standardized way to create and generate handleable errors LlamaStackError which includes a private factory function to produce FastAPI HTTPException objects that can easily be returned.
5. Creates a number of new named error types to the llama stack API that can be easily imported and handled by a client, or used to generate standardized error returns in the server.

Test Plan

The content of these changes are largely tests :)

[mergify]

This pull request has merge conflicts that must be resolved before it can be merged. @iamemilio please rebase it. https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork