Clang-tidy CI test: bump version from 15 to 16

[lucafedeli88]

This PR bumps the version used for clang-tidy CI tests from 15 to 16.
It also addresses all the issues found with the upgraded tool.

Most of the issues are related to this newly introduced check:

• cppcoreguidelines-avoid-const-or-ref-data-members
  The check enforces CppCoreGuidelines about constant and reference data members .
  In general, I understand the argument against using constant or reference data members.
  There is however one case in which I am not fully convinced by the suggestion of the tool: in PML.H const amrex::BoxArray& m_grids; becomes amrex::BoxArray m_grids; and I am wondering if this can be an issue for performances. Maybe we could consider using a (possibly smart) pointer to the BoxArray, instead of making a copy. (we are now using a pointer for amrex::BoxArray m_grids;).

Few issues were instead related to these checks:

• modernize-loop-convert
  This check was already enabled, but clang-tidy-16 broadens its scope with respect to previous versions.

• modernize-use-auto
  Only one case. I am a bit confused because this should have been found also by version 15 of the tool.

• misc-use-anonymous-namespace
  This is a new check. Actually, the issues found with this check were false positives, but they disappeared when I properly set misc-use-anonymous-namespace.HeaderFileExtensions in the .clang-tidy configuration file to recognize .H files as headers.

• misc-misplaced-const
  Only one case. I am a bit confused because this should have been found also by version 15 of the tool.

• readability-misleading-indentation [NOW DISABLED DUE TO FALSE POSITIVES]
  This check was already enabled. However, with this newer version of the tool, it seems to me that it generates some false positives. Therefore, I would like to propose to disable it. We may try to re-enable it when we will bump the version from 16 to 17.

[EZoni]

@lucafedeli88

Is there a specific reason why you want to move to version 16 as opposed to even newer versions?

[lucafedeli88]

@lucafedeli88

Is there a specific reason why you want to move to version 16 as opposed to even newer versions?

@EZoni , the main reason is to split into two PRs the changes that will be required with newer versions of clang-tidy . This is also motivated by the fact that clang-tidy-17 introduces a fair amount of new checks ! I have already opened a (WIP) PR to upgrade from 16 to 17 ( #5600 ). I expect that it will be finished by next week.
However, if you prefer a single PR, I can bump the version to 17 and address the issues that clang-tidy-17 will find.

[ax3l]

Why not simply a ref?

Suggested change

	amrex::BoxArray m_grids;
	amrex::BoxArray & m_grids;

[lucafedeli88]

@ax3l , the goal of avoid-const-or-ref-data-members is to avoid using const or ref data members, on the grounds that

Having such members is rarely useful, and makes the class only copy-constructible but not copy-assignable

This recommendation is part of the C++ Core Guidelines, and clant-tidy documentation suggests also this additional reading on the topic.

That said, we are not forced to enforce this rule in WarpX (although I think that this rule makes sense in most cases), or we could make an exception specifically for this line. For the moment I've implemented @WeiqunZhang 's suggestion, but I can revert it and just add a NOLINT.

[ax3l]

Agreed, but my suggestion just adds & not const back :D

[ax3l]

LGTM 👍