[Feature Request] Allow configuring multiple LLM endpoints with priority and automatic failover.

Description

This would allow users to use a more powerful Ollama server when available, while automatically falling back to another server if the preferred endpoint is offline or unreachable.

Example use case:

• Primary endpoint: a personal PC with a powerful GPU, only available when powered on.
• Fallback endpoint: a dedicated server with a less powerful GPU, always available.

This would allow Paperless to benefit from available hardware without requiring an external proxy or manual configuration changes.

Other

_No response_