Mir.Intrinsics functions use SimState when SymGlobalState/IntrinsicTypes would suffice

[RyanGlScott]

While working on the Rust SAW backend (see GaloisInc/saw-script#1859), I discovered that the cruible-mir memory model–related functions in Mir.Intrinsics have types that are too specific for SAW's needs. Here is one example of such a type:

crucible/crucible-mir/src/Mir/Intrinsics.hs

Lines 1139 to 1145 in df3297b

	writeRefRoot :: forall p sym bak ext rtp f a tp.
	(IsSymBackend sym bak) =>
	SimState p sym ext rtp f a ->
	bak ->
	MirReferenceRoot sym tp ->
	RegValue sym tp ->
	MuxLeafT sym IO (SimState p sym ext rtp f a)

SAW needs to write to references as part of the implementation of some of its MIR-related commands, and it ends up calling this function at some point. Or, at the very least, it should call this function, but there is a catch: it requires a SimState as an argument. This is inconvenient for SAW's needs, as the setup-related code that it uses to install MIR specifications doesn't have easy access to a SimState. (It does have access to a SimState when actually simulating the code, but that happens much later in the pipeline.)

What's more, I don't think functions like writeRefRoot even need to use SimState. If you look at the function's implementation:

crucible/crucible-mir/src/Mir/Intrinsics.hs

Lines 1146 to 1154 in df3297b

	writeRefRoot s bak (RefCell_RefRoot rc) v = do
	let sym = backendGetSym bak
	let iTypes = ctxIntrinsicTypes $ s^.stateContext
	let gs = s ^. stateTree . actFrame . gpGlobals
	let tpr = refType rc
	let f p a b = liftIO $ muxRegForType sym iTypes tpr p a b
	let oldv = lookupRef rc gs
	newv <- leafUpdatePartExpr bak f v oldv
	return $ s & stateTree . actFrame . gpGlobals %~ updateRef rc newv

The only parts of the SimState that the function ever uses are the SymGlobalState and IntrinsicTypes. Indeed, this is a pattern that holds for nearly every function in Mir.Intrinsics that currently uses SimState. We could just as well refactor writeRefRoot to have this type signature:

writeRefRoot :: forall sym bak tp.
    (IsSymBackend sym bak) =>
    bak ->
    SymGlobalState sym -> -- new
    IntrinsicTypes sym -> -- new
    MirReferenceRoot sym tp ->
    RegValue sym tp ->
    MuxLeafT sym IO (SymGlobalState sym) -- new

If we did this, then it would be much easier for SAW to call into the crucible-mir memory model.

[RyanGlScott]

I added a small number of *IO functions in #1106 for a particular SAW patch, but now I need more for another patch. Perhaps this is a sign that I should bite the bullet and add *IO entrypoints for everything in Mir.Intrinsics (or at least everything used in execMirStmt).