Add to source files compiled with no-fast-math

[ndgrigorian]

This pull request adds where.cpp, copy_and_cast_usm_to_usm.cpp, and boolean reductions sources to the source files to be compiled with -fno-fast-math. This resolves issues with floating point NaNs being interpreted as false when cast to bool with a CUDA backend.

Additionally, factors boolean_reductions.cpp and boolean_reductions.hpp the reductions submodule.

• Have you provided a meaningful PR description?
• Have you added a test, reproducer or referred to an issue with a reproducer?
• Have you tested your changes locally for CPU and GPU devices?
• Have you made sure that new changes do not introduce compiler warnings?
• Have you checked performance impact of proposed changes?
• If this PR is a work in progress, are you opening the PR as a draft?

[github-actions]

Deleted rendered PR docs from intelpython.github.com/dpctl, latest should be updated shortly. 🤞

[coveralls]

coverage: 91.145%. remained the same
when pulling ae73692 on add-no-fast-math-sources
into c578614 on master.

[github-actions]

Array API standard conformance tests for dpctl=0.15.1dev3=py310h15de555_90 ran successfully.
Passed: 908
Failed: 1
Skipped: 86

[github-actions]

Array API standard conformance tests for dpctl=0.15.1dev3=py310h15de555_90 ran successfully.
Passed: 908
Failed: 1
Skipped: 86

[ndgrigorian]

Examples of fixed functionality

In [1]: import dpctl.tensor as dpt

In [2]: x = dpt.full(10, dpt.nan, dtype="f4", device="cuda")

In [3]: dpt.astype(x, "?")
Out[3]:
usm_ndarray([ True,  True,  True,  True,  True,  True,  True,  True,
              True,  True])

In [4]: dpt.any(x)
Out[4]: usm_ndarray(True)

In [5]: dpt.where(x, dpt.asarray(1), dpt.asarray(0))
Out[5]: usm_ndarray([1, 1, 1, 1, 1, 1, 1, 1, 1, 1])

[oleksandr-pavlyk]

Thank you for tracking the issue down and fixing it @ndgrigorian

LGTM!