[FTS] compaction should be controlled/managed/herd'ed across pindexes

Currently, we use forestdb auto-compaction, which could have a downside that the multiple pindexes that comprise a single index might all "wake up" and decide to compact, all at the side time (since they all had the same creation time).

Instead, a higher level system should orchestrate compaction across multiple pindexes.

A super higher level system might even attempt that across nodes.