Const expr array length

[mohammadfawaz]

Closes #28676

This PR allows array lengths to be arbitrary expressions. Of course, we still want those expression to evaluate to a literal at compile time. We error out otherwise. Due to monomorphization and loop unrolling, type checking now has to be done multiple times which is why I had to add it, along with symbol table creation, to const_propagation_and_unrolling.rs. This seems to work fine except for a few small issues that I had to work around such as making sure we're not emitting duplicate warnings.

The default length type is u32. That is, if the length provided is an unsuffixed literal, we will assume it's a u32.

Because types can now have expressions in them (this will also be the case when we add const generic structs), I had to add a TypeVisitor and TypeReconstructor and address types where appropriate. Most of the changes happen in early passes since later passes can assume that types have been resolved (so no need to visit their expressions at that point).

The interpreter now treats const generics as regular inputs so that calling generic inline now works fine. For that, I also had to move ProcessingScript until after ConstPropagationAndUnrolling. Otherwise, we were sometimes error out in the build stage of leo test because we were removing scripts before monomorphizing.

Unsuffixied literal now get converted to type literals in const propagation when possible.

[mikebenfield]

Overall looks great.

[mikebenfield]

LGTM!