Add: MultiFabRegister

[ax3l]

Add a central location to store all MultiFabs in. The ablastr::fields::MultiFabRegister provides us with a central location to allocate, initialize and exchange fields now.

This is useful to add new fields at runtime (even Python!), access them from anywhere for read & write (incl. Python), alias them from anywhere, redistribute them on load balancing by default (unless you say: no, don't do this, during allocation), and to simplify function calls that access a lot of fields (e.g., solvers just need to pass the ablastr::fields::MultiFabRegister & WarpX::m_fields around instead of including WarpX.H).

New Headers

New ABLASTR header:

#include <ablastr/fields/MultiFabRegister.H>

New Type Aliases

using ScalarField = amrex::MultiFab *;
using VectorField = std::array<amrex::MultiFab *, 3>;
using MultiLevelScalarField = amrex::Vector<ScalarField>;

and const variants.

New Functions

MultiFabRegister provides calls to allocate & initialize (alloc_init, alias_init), test (has), access (get, get_mr_levels, get_alldirs, get_mr_levels_alldirs), list (list), remove (erase, clear_level), and remake on AMR/load balance (remake_level).
Please see the Doxygen page for all details.

These functions can be used both with warpx::fields::FieldType enums (preferred) as well as std::string (e.g., for purely runtime logic or Python hackery).

TODO

• make sure all test pass
• write more extended PR description
• Fields.H: warpx::fields::FieldType add per-enum-entry inline doc string
• use throw logic on get with missing keys
• cleanup

[EZoni]

Since there is a Direction available in AMReX, which should serve the same purpose, maybe we can try to use that one and see if it works for us. AMReX's direction is defined here:

enum class Direction : int { AMREX_D_DECL(x = 0, y = 1, z = 2) };

This could be another separate follow-up issue/PR.

[ax3l]

Yes, see the @todo in the Doxygen comment block above.

We currently need to add some implicit casts and comparison operators, thus we cannot use it.

[ax3l]

Oh ha, clang-tidy is not a required test it seems.

[ax3l]

Settings updated and fix in #5309