Build on Occupied cores when we have enough bitfields

[coriolinus]

I suggest we file the whole Occupied case for later. The exact logic that we want to employ here is going to need design of its own, because we won't want collators to wait for bitfields as a heuristic as to whether they should create a new block or not. The reason for that is that we want backing validators to have as much time to validate as possible.

AFAIK the CandidateBacking subsystem doesn't handle the non-Scheduled case anyway

Originally posted by @rphmeier in #1557

[burdges]

Yes.

We should discuss the opposite direction eventually too: A backing validator or relay chain block producers might apply roughly a parachain fork choice rule by not excluding or even just selecting a block with more descendants. Are forks limited to 1-2 block length differences or something? I suppose no..