Implement retries into WS and HTTP RPC

Right now, when we connect to WS(s) or HTTP(s) RPC, it only tries once. We should allow for retries before throwing completely.

Additionally, being able to retry the connection if we lose it would be helpful as well.

• Implement a queueing wrapper around StreamingSocket with a non-blocking send function
• Implement reconnects that send messages from the queue via the new connection as soon as established
• Expose connected state (connecting, connected, disconnected) as a ValueAndUpdates field (Because we can. It is a fair API, but not strictly required for the product)
• Make it possible for higher level applications to react to reconnects via a simple reconnected callback
• Automatically reconnect on error/completion
• Automatically re-subscribe in tendermint-rpc
• Automatically re-subscribe in ethereum

[webmaster128]

Good point @ethanfrey

To break this down into smaller steps, I suggest

1. Implement a queueing wrapper around StreamingSocket with a non-blocking send function
2. Implement reconnects that send messages from the queue via the new connection as soon as established
3. Expose connected state (connecting, connected, disconnected) as a ValueAndUpdates field (Because we can. It is a fair API, but not strictly required for the product)
4. Make it possible for higher level applications to react to reconnects via a simple reconnected callback
5. Automatically re-subscribe in tendermint-rpc
6. Automatically re-subscribe in ethereum

[ethanfrey]

I agree with the above list. I think it still has some utility even in smaller chunks, maybe a series of PRs.