Draft: Scheduler: avoid instantiating Tasks that the host_architecture does not match (related: #322)

Make scheduler filter out WorkRequests where the host_architecture of the worker and the host_architecture of the WorkRequest.data does not match.

This should fix #322 (closed) (I haven't tested it end to end).

Avoids instantiating each Task to avoid the problem mentioned in #322 (comment 470184)

Having said that: I am not sure that right now is the good time to optimise the scheduling process (but, someone else, might say that is the good time to not make it worse).

I agree with the problem, the optimisation might not be in the right place (and could be done earlier I think changing the scheduler approach more).

What I want to say: this is better than before, should fix the problem; it's not perfect but we don't have a scheduler performance just yet.

I think that I might have spotted some big mistakes of the schedule tests (not in the new code, in older code. I will comment on that tomorrow).