[CodeGen] Prepare for introduction of v3 and v5 MVTs

AMDGPU would like to have MVTs for v3i32, v3f32, v5i32, v5f32. This
commit does not add them, but makes preparatory changes:

* Exclude non-legal non-power-of-2 vector types from ComputeRegisterProp
  mechanism in TargetLoweringBase::getTypeConversion.

* Cope with SETCC and VSELECT for odd-width i1 vector when the other
  vectors are legal type.

Some of this patch is from Matt Arsenault, also of AMD.

Differential Revision: https://reviews.llvm.org/D58899

Change-Id: Ib5f23377dbef511be3a936211a0b9f94e46331f8
llvm-svn: 356350