[OPENMP, NVPTX] Fixes for NVPTX RTL
Patch fixes several problems in the implementation of NVPTX RTL.
- Detection of the last iteration for loops with static scheduling, no chunks.
- Fixes reductions for the serialized parallel constructs.
- Fixes handling of the barriers.
Reviewed By: grokos
Subscribers: Hahnfeld, guansong, openmp-commits
Differential Revision: https://reviews.llvm.org/D48480