Resubmit: Adding parallel support for the LLVM backend. (#54122)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/54122
Test Plan:
* USE_TBB=1 ATEN_THREADING=TBB python setup.py develop --cmake
* USE_TBB=1 ATEN_THREADING=NATIVE python setup.py develop --cmake
* USE_TBB=1 ATEN_THREADING=OMP python setup.py develop --cmake
* cd build; ninja bin/tensorexpr_bench
* bin/test_tensorexpr --gtest_filter="*Parallel*"
Reviewed By: bertmaher
Differential Revision: D27109802
Pulled By: zheng-xq
fbshipit-source-id: db159466d0b46357bcf0fbefb36094bee312368c