Expression partitioning to respect scope validity

[gatesn]

Currently the expression partitioner will not correctly respect the validity of the scope array.

This is blocked on a mask expression existing: #1900

[gatesn]

Is this true? Or would any expression that altered the result array validity already be part of the root expr, and therefore respected?

Maybe the problem is that we need to ensure that when getting the child field of a layout, that the parent's validity is pushed down. We could enforce this at write time.

[robert3005]

We assert that structs when being written don't have top level nullability