Improve Fallback Logic for Endpoints

[LeoPatOZ]

Currently all providers are tried in order when calling regular rpc endpoints. This means that even if a provider is down (e.g the primary provider) we will still try to call it then failover to the next provider (which might also be down). This can cause significant lag / degradation in performance as you need to cycle through known dead providers before getting to the one that works.

Ideally there's global state tracking of which provider is working across rpc calls. This is similar logic to the robust subscription.

We should share the same logic between the robust subscription and regular calls.