2nd follow up from #16073 is to
have Astrid use the same flow for backward block downloads as Ethereum -
this unites the 2

benefits of this are:
- the new backward block downloader abstraction matures more - I found 2
small issues with it while doing this work that are now fixed - also
it's interface is more flexible to cater for both use cases now
- there will be performance gains in astrid when handling milestone
mismatches - the new downloader is much quicker to figure out what peers
it can backward download the milestone from and gives a fail/success
answer much quicker (no 60 sec stalling when we hit a peer that doesn't
have the milestone as previously) - also when getting a new block hash
event we can directly backward download to a connection point in the
canonical chain builder (instead of first downloading 1 block from a
peer and then realising that we need to download more because there is a
gap and sending more requests)