Failover on RPC node maintenance

[mschneider]

We noticed an issue while running the aggregator, that keeps causing maintenance on our end. Basically the websocket might get locked in a weird state, when the RPC endpoint goes down for maintenance. The error looked like this:

ws error: connect ECONNREFUSED xxx:443
ws error: connect ECONNREFUSED xxx:443
ws error: connect ECONNREFUSED xxx:443

At some point the socket stopped trying to reconnect and I had to manually restart the oracle to make it work again.

Ideally the code would just allow us to specify a preference of rpc nodes and switch to a different one, in case the primary one becomes unavailable. Also it should switch back, when the primary one becomes available and is ready to accept transactions Could you look into that?