When no allocator passed to the coroutine:

Schedule bulk work of size `__count` on parallel scheduler, calling `__rcvr` for different subranges of [0, __count), and using `__scratch` for preall

Schedule bulk work of size `__count` on parallel scheduler, calling `__rcvr` for each item, and using `__scratch` for preallocated memory.

NOTE: __rcvr's type is unconstrained because the thing we pass doesn't satisfy receiver

Ensure that `__storage` is aligned to `__alignment`. Shrinks the storage, if needed, to match desired alignment.