Bugfixes

[jngrad]

Description of changes:

• recalculate up-to-date particle box images in the python interface and hdf5 writer
  • fixes #4940
• address propagation flags regressions in the checkpointing and virtual sites codes
  • fixes #4935
  • fixes #4936
• update LB pdf/force/velocity fields together to avoid out-of-date cached velocities
  • fixes #4934
  • add unit tests for the LB field accessors (both C++ and CUDA)
• extract the code coverage logic to a self-contained file and forward coverage compiler flags to Ccache
  • closes #4943
  • developers can now reproduce Codecov coverage reports locally
• add support for all CMake name boolean constants in the pypresso launcher
  • users are no longer required to use ON/OFF when configuring boolean variables of the CMake project
• improve documentation of thermostats and propagation flags
  • fixes #4930

[RudolfWeeber]

I reviewed the commits 1-3 and 6-7. They can be merged.

[RudolfWeeber]

For 3 and 4 it would be helpful to know in broad terms what the actual cahgnes are. That would make it clearer to me what to look for in the diff

[jngrad]

For 3 and 4 it would be helpful to know in broad terms what the actual cahgnes are. That would make it clearer to me what to look for in the diff

• commit 3: it is easier to look at the *.tmpl.{h,cu,cuh} template files. The pdf/vel/force files can no longer be edited independently from one another, since updating any one of them requires updating at least one of the other two. I also implemented a few missing slice accessors.
• commit 4: it is bad practice to declare CMake functions without a prefix like espresso_, because we can accidentally overshadow an existing CMake function and run into nasty undocumented behavior, e.g. infinite recursion. See Craig Scott's Do Not Redefine CMake Commands blog post for more details (he's a CMake maintainer). A function called unit_test() doesn't exist in CMake yet, but it's not a far-fetched idea that such a function name could be introduced in a later release of CMake, or in one of its utility modules. All other ESPResSo-specific CMake functions had been renamed in older refactoring PRs. This one was renamed only now because it was quite complex and required CMake 3.25 functionality to be rewritten in a clean way. Also, it used to take multiple source code files as input, but it was totally unclear from the code how it was supposed to link them together, especially if they were mixing CUDA and C++ files. Now it only takes 1 source file a input, from which the name is deduced. So we can no longer have a fantasy test name that is unrelated to the source file name.
• commit 5: when running cmake .. -D ESPRESSO_BUILD_WITH_UBSAN=YES, the build is successful but pypresso fails to launch with a cryptic error message that took me 30 min to figure out. The pypresso launcher was too fragile and didn't consider that other keywords like true/yes/y/on and other less common named constants can evaluate to true.

[reinaual]

I've checked commit CMake: Rewrite unit_test() framework and Shell scripts hardening, they are both fine.