add inttype, floattype like uinttype

[JeffreySarnoff]

uinttype is an unexported function that maps IEEEFloat types to their bitwidth-matched Unsigned types.
This PR adds inttype and floattype (also unexported). inttype maps IEEEFloats to bitwidth-matched Signed types and floattype maps uinttype(T), inttype(T) where {T<:IEEEFloat} to the bitwidth-matched IEEEFloat type.

I use these functions often -- less elegantly defined (overloading reinterpret or rem). As they are unexported, I did not add NEWS. Since there are no tests for uinttype, I did not add tests .. although I am happy to do so for all three functions should that be preferred.

[kimikage]

FixedPointNumbers defines and exports floattype, which is used by packages related to graphic and image processing.
It's weird that the floattype is defined in FixedPointNumbers, and it's not bad to define it in Base. However, the floattype in FixedPointNumbers is not consistent with Base.float due to performance concerns (cf. JuliaMath/FixedPointNumbers.jl#127).

Of course, this is just the problem in a specific ecosystem.

[kimikage]

cf. #30445

[oscardssmith]

bump. I think this would be good to have (at least internally).

[kimikage]

I agree that it is not a problem because they are not exported, but I do not think it is good practice to use functions that are not exported.

BTW, we can find inttype in a surprising place. 😄

julia> Base.Threads.inttype(Float32)
Int32

The root of the problem is that whether you associate the names inttype or floatype with reinterpretation or conversion depends on the context and personal mindset.

[musm]

I've also needed these in the past. Since they're internal I find these definitions relatively harmless and are more for convenience.

Sans objections I plan on merging this in a day.