feat(ui): Utilize OpenAI for backend

[andrewrisse]

closes #528
closes #414

• Changes naming convention to use threads instead of conversations
• Refactors to use OpenAI types for threads, messages, and assistants
• Refactor to no longer save threads, messages, and assistants directly to Supabase, but instead uses OpenAI endpoints
• Bug fixes for AssistantTile text, Assistant Avatar uploading
• Adds new mocking setups for tests

Key differences with conversion to OpenAI:

• OpenAI uses strings for ids, not uuids (ex. thread_12345), this required schema changes
• OpenAI nests its message content into complex objects, but Vercel AI SDK does not. Required some type conversions and helper functions to ensure compatability
• OpenAI is slower to respond and my be rate limiting us which causes Playwright test timeout issues, refactor required to improve reliability
• Frontend is now limited to the API endpoints available with OpenAI and can no longer build it's own custom queries to pull data from Supabase. This drove several changes to basic application logic/data fetching as well as tests.
• When viewing "files changed", several files look like they were added, but it's really just a name change (ex. conversation -> thread naming convention)
• Migrations required for user profiles in Supabase - addition of "thread_ids" which tracks all of their threads so we can load them all when the app loads
• Migration to delete conversations and messages from supabase. This will result in data loss, but we have accepted this for release 0.8.0. Alternative is to download all conversations and messages from Supabase, reformat into OpenAI type conventions, then re-save to Supabase once new LeapfrogAI API is ready.

This PR is currently deployed at ai.uds.dev

[BillyFigueroa]

Big PR but great work