Code perf: optimize typeof equality in inlined code

This is is recording a potential simple to our optimizations in Optimizer.fs

The .NET CLRs (both Framework and .NET Core) have an optimization where code like

if LanguagePrimitives.PhysicalEquality typeof<'T> typeof then
...
elif LanguagePrimitives.PhysicalEquality typeof<'T> typeof then
...

is optimized if the JIT is specialising the code for a struct type.

Here is a gist showing its use from F#: https://sharplab.io/#gist:3980a4de89ed3dc1d0d86d4120df2cbc. (UNfortunately the ASM is only shown for x86 desktop, I'd like to see it on RyuJIT for .NET Core 3.0 too). The construct `(# "" x : int #)` is being used to get an unsafe value coercion (as in prim-types.fs)

We should do the same optimization in the F# compiler itself when we inline and type-specialize code. It should be a simple addition to Optimizer.fs. This will provide a hook for numerics libraries that allow instantiation on a limited set of scalar types to an important range of optimizations.