The first open checkpoint for a new active vbucket should start with id 2

There is a boundary case that caused data loss if the first checkpoint for a new active vbucket starts with id 1.

For example,
1. Node A started and has the active vb 0 with the open checkpoint 1
2 Load some items into the open checkpoint 1
3. Restart A and then A will have the open checkpoint 1 after warmup, which is empty.
4. Node B is newly joined and has the replica vb0 with the open checkpoint 1
5. Replication from A to B for vb0 doesn't schedule backfill because we thought that both node A and B have the synchronized open checkpoint for vb0.