SBliFF fork handling

[brenzi]

Under the assumption of BliFF on L1, SCV can detect and handle forks like this:

get sidechain.last_finalized_block
if head.is_lineal_descendant_of(last_finalized_block)
  //all good, no fork. continue
else
  // we're on a fork. 
  find peer who knows last_finalized_block
  fetch blocks backwards from last_finalized_block until we hit a known block in our sidechain_db
  if peer can't deliver or no match is found, kill own state and sidechain_db and request full provisioning from peer (i.e. by bruteforce deleting the state&db on disk and restarting the service)
  if successful, fetch all recent blocks from peer since last_finalized_block
   
  

prior work

Consider thoughts in here:
#687

ForkTree

There is an implementation ForkTree already which we could use here. #1196

import_or_sync_block() (in import queue worker) currently simply tries to import blocks in order of their blocknumber and purges blocks that don't match.

Instead, we should maintain a fork tree of all block candidates we know. upon a new finalization ForkTree can prune branches that are finally abandoned

design

design decisions:

• validateer fetches LatestSidechainBlockConfirmation directly, without feeding it into STF state first through event listeners, because sidechain state could stall, then we're deadlocked

tracking of finalized sidechain block

STF event

execute_top_pool_trusted_calls_internal()

ProcessBlockImportQueue.process_queue()
fetch LatestSidechainBlockConfirmation
for all competitor blocks in queue for the next sidechain block
ensure is lineal descendant of LatestSidechainBlockConfirmation

[brenzi]

will close this in favor of concrete design in #1587