llvm: Remove LLVM declares from LLVM overrides

[langston-barrett]

Fixes #1138.

While this PR adds lines overall, a large chunk of that is identity cases in pattern matches that are included for the sake of future-proofing those cases. Fundamentally, this is a significant simplification.

Some improvements to the Cast module that were enabled by this change:

• A type family that specifies the sorts of casts that are performed
• Make all functions total
• Separate casts to and from the Crucible-LLVM ABI

One breaking change: Crucible-LLVM S-expression files now have to use the exact signature of the underlying override. See the changes to tests for examples of this breakage. Not a huge deal, as the S-expression syntax is mostly developer-facing.

After this change, it would be feasible to completely remove the quasiquoters. I haven't done that in this PR, but would be happy to do so if requested. It might be nice to instead move them to the type level as quasiquoters for LLVMOverride p sym ext args ret where args and ret are determined by their LLVM types.

[langston-barrett]

Status: I'm adapting GREASE to these changes to make sure they work in practice.

[EDIT]: See GaloisInc/grease#525

[RyanGlScott]

Do you envision these changes being breaking to most downstream consumers? (I mainly ask this in a SAW context, since we are in the middle of planning new releases soon.)

After this change, it would be feasible to completely remove the quasiquoters. I haven't done that in this PR, but would be happy to do so if requested.

That is an interesting idea, although I do like the fact that the quasiquoter syntax is much more terse than what the equivalent Haskell code would be. As such, I'm a bit on the fence about this idea. Perhaps we could discuss this in a separate issue?

[langston-barrett]

Do you envision these changes being breaking to most downstream consumers?

Only the ones (like crucible-llvm-syntax or GREASE) that use the S-expression syntax. I would expect that SAW, like Crux-LLVM, will require no changes. That being said, I'd prefer to delay merging until the SAW release is complete just in case.