Better checkpoint signature process and support for batch commitment

[adlrocha]

Checkpoint signature and commitment by subnets represents a core mechanism in the operation of HC and, as such, its implementation should be battle-tested. There may be some rough edges in the current MVP implementation of the protocol that we should review, audit and, potentially, improve before moving into production. Mainly:

• What happens if we miss the signature of a checkpoint in a signing window, do we lose the cross-messages included? Cross-msgs should be batched if checkpoint signing window passes? #156
• We currently batch checkpoint commitment by sending the commitment message to the message pool in order for them to be included in the next block. We follow a "fire-and-forget" approach for the signature and submission of checkpoints but, is this enough? What if the message sent to the message pool are never included in a block, should we retry?
• The process of detecting that a new checkpoint is available to sign, and that it may submitted is automatic. Should we include a manual fallback for a "human operator" to be able to submit it manually in case it fails?
• And finally, we should figure out the gas management for checkpoints, what if all miners run out of gas? Does this affect in any way the protocol?