[PATCH] Add support for scalarizing ctlz_zero_undef

[Petar Jovanovic]

>>! In D4645#12, @chandlerc wrote:
> This seems fine (as Matt said) but don't you need to do the same for CTTZ?

I was solving a particular failure, so this is why I have not included CTTZ_ZERO_UNDEF. But, yes, it should be done for CTTZ too.

I can update the case in ScalarizeVectorResult() with CTTZ.
(Or it should be another patch with additional test case?)

http://reviews.llvm.org/D4645