[Feature]: Avoid actual chat request in model provider availability check.

[oasisfeng]

Is your feature request related to a problem?

Please don't use an actual chat completion request as the implementation method of model provider availability check. It's costly for advanced models (e.g. OpenAI o1).

The current implementation also sends chat completion request for all the models in the specified provider. This is very slow and costly if many models are enabled for that provider.

Describe the solution you'd like

Consider using just "/v1/models" to the check the availability (of base URL and API key), as it is free of charge.

Describe alternatives you've considered

No response

Additional Context

No response

[kangfenmao]

Not all service providers offer the /v1/models interface

[kangfenmao]

What are some other good suggestions?

[oasisfeng]

Probably try "/models" first, if absent, fallback to "/completion"?