add sleep after new block to the request_decider

[Jiloc]

Description

Part of : #1237

Changes

• originally part of fix: missing signers' votes #1243 , this change add a new requests_processing_delay to the request_decider. This will give additional time to the bitcoin nodes to get in sync before receiving the vote decisions from the signers.

Checklist:

• I have performed a self-review of my code
• My changes generate no new warnings
• New and existing unit tests pass locally with my changes

[djordon]

Looks good.

Let's wait to get more eyes on this before merging, in case there is some edge case that we're blanking on that would make this a bad idea.

[matteojug]

Thinking out loud: assuming everyone gets the bitcoin block at exactly the same time, with the sleep as is here we may risk having the coordinator starting the round before the decisions manage to get to it (like, everyone broadcast its own decision, then the coordinator starts but it only has its own decision).
I think this is slightly worse than before, but it could be equivalent.

[Jiloc]

How about we separate the two timeouts: one in the RequestDecider after a new block, and another in the TxCoordinator after receiving the signal from the RequestDecider? We could set them both to 15 seconds.

While this isn't an ideal or final solution, I believe it's a reasonable compromise. It should still allow us to receive the decisions and blocks in time most of the time.

#1243 will remain the final solution. And if we implement a filter to only resend messages that didn't receive all replies, we could significantly reduce the number of repeated messages.