Make TensorIterator, SparseTensorMath and UnaryOps clang-tidy clean (#55087)
Summary:
Disable `cppcoreguidelines-macro-usage` as PyTorch codebase uses a lots
of macros that violate this rule.
Disable `bugprone-reserved-identifier` and
`performance-unnecessary-value-param` as those checks are very slow
Add `NOLINT` to DEFINE_DISPATCH as it introduces non-const global variables
Replace `for(auto i = 0; i < lim; ++i)` with `for(auto i: c10::irange(lim))` throughout the modified files
Pull Request resolved: https://github.com/pytorch/pytorch/pull/55087
Reviewed By: samestep
Differential Revision: D27475822
Pulled By: malfet
fbshipit-source-id: 2651a4b3dc062066a15e69380354414a198fb279