Mesh refinement: Adding IO

[SeverinDiederichs]

This PR adds a first working version of IO of the refined level.

In the OpenPMDWriter class, the series is now a vector over levels. Each level has its own series and writes to its own sub-folder in the IO path. If only the 0th level is used, no sub-folders will be generated.

Then, the IO is now a loop over levels. Additionally, the copying of the slice to the diagnostics is now a loop over levels.

The resizing of the diagnostic arrays is a loop over levels. To not yet touch the other functions, it is done with local variables.

To avoid complications with the Poisson solver (the staging area needs to exist per level), the Poisson solver is not even allocated for other levels than lev = 0. This is handled in #527.

Additionally, the boxarray for the first level is not set correctly.

In the current version, the code crashes if used with a beam in the binning. No beam can be used with MR at this point.

• Small enough (< few 100s of lines), otherwise it should probably be split into smaller PRs
• Tested (describe the tests in the PR description)
• Runs on GPU (basic: the code compiles and run well with the new module)
• Contains an automated test (checksum and/or comparison with theory)
• Documented: all elements (classes and their members, functions, namespaces, etc.) are documented
• Constified (All that can be const is const)
• Code is clean (no unwanted comments, )
• Style and code conventions are respected at the bottom of https://github.com/Hi-PACE/hipace
• Proper label and GitHub project, if applicable

[MaxThevenet]

Looks good, thanks! See a few comments below.

[ax3l]

Awesome! :)

Let me cross-link the WarpX implementation and definitions we drafted for the openPMD standard:

• WarpX implementation: openPMD: Support for mesh refinement BLAST-WarpX/warpx#1842
• Standard definition: Ext: PatchBasedMeshRefinement openPMD/openPMD-standard#252

Note that the WarpX implementation currently still over-allocates file size for the refined level for HDF5 (we use this with ADIOS2 at the moment). A follow-up PR tuning HDF5 further would only declare a smaller box per patch on the refined level to work-around this, as noted in both links :)