Improve Sesame voice chat compatibility with current web flow and auth bootstrap

[SaseQ]

Summary

This PR updates the client to match current Sesame web behavior and fixes connection failures seen with the old flow.

What changed

• Updated WebSocket defaults to align with web app handshake:
  • client_name default is now Consumer-Web-App
  • Origin default is now https://app.sesame.com
• Added connection diagnostics in WebSocket layer:
  • captures and exposes server-side error messages (last_error)
  • unblocks connect wait on explicit error/close paths
• Fixed call setup payload:
  • call_connect now sends settings={"character": <character>} by default
  • optional preset support kept via settings={"preset": <preset>} when provided
• Added backend user bootstrap API call:
  • SesameAI.ensure_external_user(...) calls PUT /api/external/user
  • invoked during voice chat authentication so fresh anonymous tokens can be used without manual setup
• Extended token handling utilities:
  • TokenManager.set_tokens(...)
  • TokenManager.load_firebase_auth_user(...)
  • get_valid_token(..., allow_anonymous=True)
• Improved CLI auth options in examples/voice_chat.py:
  • --id-token
  • --refresh-token
  • --firebase-auth-json
  • --no-anonymous
  • --client-name
• Updated README to document the new auth/connect behavior and options.

Why

Sesame backend behavior changed compared to the original wrapper assumptions. The old flow caused failures such as:

• 403 Access denied. Anonymous users not allowed
• 5002 Settings were not found
• 404 No user found

This PR aligns the Python client with the currently working web flow and restores a smooth "just run the example" experience.

Scope

Modified files:

• sesame_ai/websocket.py
• sesame_ai/api.py
• sesame_ai/token_manager.py
• examples/voice_chat.py
• README.md

[matt-norris]

looks like this projects been abandoned but nice work