Remove `EndpointPickerNotHealthy` condition from InferencePool Status

[ahg-g]

What would you like to be added:

Remove EndpointPickerNotHealthy condition from InferencePool Status

Why is this needed:

InferencePool Status is typically managed by the gateway controller, which in the normal circumstances wouldn't monitor the epp health status, the proxy does, and so in most cases it will not be practical for the gateway controller to set it

/cc @danehans @robscott

[kfswain]

How will we signal poor Pool health?

[ahg-g]

Good point, so the tricky part is that not all gateway implementations have access to this signal to actually report it, and so it seems worse to define the condition, but it is mostly not implemented. I think it is reasonable to align with the capabilities of the the initial set of gateway controllers that aim to implement the API and expand later.