Implement snapshot creation policy

The current implementation of creating a snapshot is only executed after `handle_committed_normal_entry`.

I think there could be better ways.

For example, we can implement the following policies:

Create a snapshot when the size of the FSM exceeds a specific threshold.
Generate a snapshot when the number of log entries surpasses a certain count.

Note: `snapshot` contains several metadata configuring the cluster, so wrong implementation of this would break the consistency of the raft loop.

Start by tracing snapshot creation from handle_committed_normal_entry and reviewing the snapshot metadata used by the raft loop. Compare the proposed FSM-size and log-entry-count triggers, then clarify which policy and thresholds are required. Done means snapshot timing is defined without breaking cluster metadata or raft-loop consistency.

Written by the indexing model from the issue text.