ReplaceWithVeclib.cpp would construct overload name using all the arguments in the intrinsic, but overloads should only be constructed from arguments for which isVectorIntrinsicWithOverloadTypeAtArg returns true, including the return (-1).

Fixes llvm#111093

If isVectorIntrinsicWithScalarOpAtArg() returns true for an argument or return, we should not attempt to modify the type or use the element count, even if it is a vector.  This allows the op to be identified as something that doesn't vectorize/scalarize with the width of the intrinsic, but remains the same, even if it is a vector argument.

Initial ElementCount will only be the return element count if it's a vector and isVectorIntrinsicWithScalarOpAtArg returns false.

This also returns the control flow in the loop to be closer to the original.

Remove inconsistent comment about VFABI return type assumption, since we have the accessors that tell us how to handle it for the operation.

Fix formatting.