Limit how often the mempool revalidates its contents during bulk sync

The mempool currently revalidates its contents whenever the current ledger state's tip point changes. During bulk sync, that happens ~hundreds of times per second. So if the mempool were somehow significantly non-empty, that could be costly overhead.

In contrast, once the node has caught up, the ledger tip should instead only evolve once per several seconds (on mainnet, at least).

1) We're not sure it would be worthwhile to rate-limit this overhead, but 2) it seems like it could indeed be important to do so. So it's worth asking the question this Issue represents. EG If the mempool is large, could revalidating it for each new block (~3 times per minutes when caught up) be "too often"?