Add health check action

[TanGentleman]

• Uses primary model name to validate endpoint

The ChatOpenAI object looks like this:

ChatOpenAI(client=[...], openai_api_base='http://localhost:4000/v1')
>>> a.openai_api_base
'http://localhost:4000/v1'

For supported endpoints, we'd ping the server to ensure healthy connection before downstream tasks.

The task action list would be something like: [create_action(ActionType.HealthCheck), [...]]

[TanGentleman]

I've implemented a version of this, though I'm running into a bug where the latency is very high!

Even for the LiteLLM proxy I'm hosting locally — the regular health check runs instantly, but within the flow of my graph, it's causing a huge amount of lag before executing the API call. I'll test with sync/asynchronous quirks in the future, as this is a low-priority feature