Previously, `apply_block_relevant` used `batch_insert_relevant` which
allows inserting non-topologically-ordered transactions. However,
transactions from blocks are always ordered, so we can avoid looping
through block transactions twice (as done in `batch_insert_relevant`).
Additionally, `apply_block_relevant` now takes in a reference to a
`Block` instead of consuming the `Block`. This makes sense as typically
very few of the transactions in the block are inserted.
a7d01dc39a