Change clang-format PointerAlignment to Left

[upsj]

Not sure if this is controversial, but following a previous discussion in Slack, it sounded like this might be a sensible choice now that we pushed a release.

This avoids the clang-format inconsistencies related to variadic pointer and reference parameters with PointerAlignment: Right.
Older clang-format versions format T&&... a as T &&... a, but newer versions format it as T &&...a, while normal
variadic parameters are formatted as T... a any time. This inconsistency is not present for PointerAlignment: Left
Also, ignoring multiple variables being declared in the same expression (which we don't do anyways), type* obj makes it slightly more clear that the left part is the type and the right part is the variable being declared, while type *obj puts more emphasis on what obj is pointing on.

Related discussion of the behavior differences: https://stackoverflow.com/questions/65964384/how-to-make-clang-format-11-put-a-space-after-variadic-template-decleration
Discussion on attachment in general: https://www.stroustrup.com/bs_faq2.html#whitespace

[codecov]

Codecov Report

Merging #872 (1f476bd) into develop (4046ebc) will increase coverage by 0.01%.
The diff coverage is 90.33%.

@@             Coverage Diff             @@
##           develop     #872      +/-   ##
===========================================
+ Coverage    94.27%   94.29%   +0.01%     
===========================================
  Files          423      423              
  Lines        34661    34659       -2     
===========================================
+ Hits         32677    32680       +3     
+ Misses        1984     1979       -5     

Impacted Files	Coverage Δ
common/unified/components/precision_conversion.cpp	100.00% <ø> (ø)
common/unified/matrix/coo_kernels.cpp	100.00% <ø> (ø)
common/unified/matrix/csr_kernels.cpp	100.00% <ø> (ø)
common/unified/matrix/dense_kernels.cpp	100.00% <ø> (ø)
common/unified/matrix/diagonal_kernels.cpp	92.85% <ø> (ø)
common/unified/preconditioner/jacobi_kernels.cpp	31.81% <ø> (ø)
common/unified/solver/bicg_kernels.cpp	100.00% <ø> (ø)
common/unified/solver/bicgstab_kernels.cpp	100.00% <ø> (ø)
common/unified/solver/cg_kernels.cpp	100.00% <ø> (ø)
common/unified/solver/cgs_kernels.cpp	95.45% <ø> (ø)
... and 130 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4046ebc...1f476bd. Read the comment docs.

[yhmtsai]

I am fine with this change because we also have a rule to make the declaration clear.
In short, we do not allow to declare two variables in the same line. we must keep forcing it
Thus, we do not have the confusion introduced by pointer left alignment.
Others changes are introduced by make format with new .clang-format, right?

[upsj]

I ran clang-format -i on all files matching *.cuh,*.cu,*.hpp.inc,*.cpp,*.hpp manually, since make format only picks up changed files. Anything I might have missed?

[yhmtsai]

will this be two lines?

[sonarcloud]

SonarCloud Quality Gate failed.   

0 Bugs
0 Vulnerabilities
0 Security Hotspots
2 Code Smells

20.0% Coverage
40.5% Duplication

[Slaedr]

This might significantly complicate the batch-develop workflow now...

(Also: const int* const x looks weird compared to const int *const x mainly because the second const applies the the pointer, and the second one suits that better. 😅 )