[FTS] rebalance of replica pindexes might skip the "wait until caught up" phase

Realized this from diagnosis of MB-18315...

MCP does "wait for catch up" maneuver for "moving" pindexes whether they are either primary and replica pindexes, so there's a possible optimization to skip this "wait for catch up" for replicas (only doing the catch-up waiting for primary pindexes only).