Snapshots should be generated at regular interval

[jumaffre]

Just like signature transactions, snapshots should be generated at regular interval (either elapsed time or number of signatures).

The hash of the snapshot should be written to the (new) ccf.snapshots table so that the snapshot becomes endorsed by the leader node that has generated it.

Once generated, snapshots should be stored on disk and retrievable for when a new node joins the network.

• 1. Generate snapshots on primary node at regular interval (based on a cchost CLI parameter)
• 2. Store snapshot evidence in new ccf.snapshot_hash table
• 3. Make snapshot generation asynchronous (snapshot generation takes time and it shouldn't block the thread on which it is started on)

[jumaffre]

I wonder how this intersect with the ledger chunking feature. On recovery, we will probably want to only pass the latest snapshot + the ledger suffix (i.e. ledger after the snapshot) to the recovery node. Perhaps a new ledger chunk could be created when a new snapshot is generated?