Split schedule many

[marioskogias]

Attempt to refactor schedule many as proposed in #49

[mjp41]

So I think Vishal is right. So there are two cases where we need to specify that the read_available flag:

• If the queue was empty, i.e. exchange returned nullptr. https://github.com/microsoft/verona-rt/blob/45e93715955eaca6ad7ede0740c544bc5eaed9b6/src/rt/sched/behaviourcore.h#L882
• If the previous slot had read_available set. https://github.com/microsoft/verona-rt/blob/45e93715955eaca6ad7ede0740c544bc5eaed9b6/src/rt/sched/behaviourcore.h#L943

The second of these is after the release phase, so I think could cause a lost wake up. The first one is in the correct place. I think we need to flag the read_available for read only chains in:

https://github.com/microsoft/verona-rt/blob/45e93715955eaca6ad7ede0740c544bc5eaed9b6/src/rt/sched/behaviourcore.h#L849-L853

[vishalgupta97]

The read reference count also needs to be taken along with marking the slot as being read available before marking the slot as ready otherwise read and writes can happen in parallel.

[marioskogias]

I am not sure I fully understand the concern. I see the difference before and after the refactoring, but it is not clear to me why in this example A will not wake up B. Is the problem that B will never be scheduled or that it will not be scheduled as early as it could?

I think that given that we call resolve() at the very end of the function, none of the behaviours are run before this function finishes.