feat(block-producer): detect and abort on chain desync

[Mirko-von-Leipzig]

This detects situations where the block-producer disagrees with the store on the current chain tip. This should only be possible under rare circumstances, but one possibility is:

• block-producer submits block N+1 to store via gRPC
• store processes and persists block N+1
• block-producer cancels its gRPC request e.g. due to timeout
• block-producer rolls back local state to N
• block production spins endlessly as block N+1 is now rejected by store

There are multiple points to detect this. This code determines by comparing the block number being built against the input's header returned from the store. The block builder task panics, which should abort the block-producer process.

Note that this error isn't really recoverable; and requires a full reset of the block-producer. Aborting and restarting is the easiest way to implement this.

Closes #1513.

[bobbinth]

Looks good! Thank you! I left a couple of small comments inline.