Improve web3 reconnection logic in offchain

[monitz87]

Right now offchain sets no timeout between reconnection attempts. Not only is this pretty intensive in CPU and memory, it may also cause reconnection to fail (since it doesn't give any time for the socket to check if the connection was properly established before pinging again). It may also be attempting multiple reconnections at once (since there are 3 listeners). This may have to be revisited and improved, but there is a pending PR in web3 that handles all of this (including resubscribing to events)

The conflicting scenario arises when you start the local blockchain, start offchain and then shut the local blockchain down.