fix(tree): remove_blocks fixes, return hash and number in persistence task

[Rjected]

This does many things:

• Make remove_blocks_above and save_blocks output both a block number and hash.
• Make remove_until take an upper bound block number, a last_persisted_block_num and hash, and an optional finalized_hash.
• remove_until will now make sure that it only removes canonical in-memory blocks if they are in the same chain as the last persisted block. This was motivated by hive tests where the persistence task would run in between a reorg. In this case, we were removing canonical blocks that were not actually persisted yet.
• remove_until is split into two methods, remove_canonical_until and prune_finalized_sidechains.
• is_canonical is introduced as a helper for remove_until
• remove_persisted_blocks is modified to only remove blocks if they are actually part of the persisted chain.
• Blocks that have been reorged, but are not on-disk, are reinserted into the tree.
• remove_blocks_above was incorrect, removing exclusively for some pieces of data, and inclusively for others. This is fixed.
• There is now a VecDeque in case there are multiple reorgs in a row.

closes #10570

[mattsse]

only nits

I was able to follow and the vecque for queued disk reorgs makes sense