Energy-preserving explicit PIC-MCC algorithm: electromagnetic

[EZoni]

Overview

Follow up on #5955 and implement the electromagnetic part of the energy-preserving PIC-MCC algorithm described in (Vay, Jean-Luc, et al. "Energy-preserving coupling of explicit particle-in-cell with Monte Carlo collisions." Physical Review E 111.2 (2025): 025306) (i.e., explicit PIC coupled with Monte Carlo collisions).

Notes

• Closes Implement Energy-Preserving PIC-MCC Loop #5851.
• Production-ready implementation of [WIP] Splitting x- push #4866, [WIP] mcc in mid x push #4867, [WIP] Mid x push with exact diags #6302 (without diags change).
• Alternative algorithm for the one in [WIP] Split of momentum push for explicit solve with insertion of collision… #4726.

To do

• Add a new electromagnetic test.
• Implement the new algorithm.
• Test & debug:
  • Fix GPU issue that seems to be resolved by passing communicate=1 to AddRealComp.
  • Fix charge deposition (deposit old $\rho$ in the first half, new $\rho$ in the second half).
  • Check tests that fail and reset the checksums if needed.

[EZoni]

Basically the goal here will be to have the analysis of the test test_2d_collisions_split_position_push_electromagnetic pass without breaking any other tests, with respect to this CI reference starting point: https://dev.azure.com/BLAST-WarpX/WarpX/_build/results?buildId=2275.

[EZoni]

Discussed with @oshapoval offline:

• The momenta that are averaged should be the momentum at the end of the first position half-push and the momentum after the collisions.
• The relative time used for the current deposition should not need any correction.

[EZoni]

@RemiLehe @oshapoval

I start requesting your review in case you have comments on the early draft. I don't think it's working correctly yet.

[oshapoval]

For better performance, reproduce the logic in HandleParticlesAtBoundaries to ensure only local communication.

[EZoni]

Maybe this can be a follow-up optimization PR, since I think it'd require some refactoring of the code in HandleParticlesAtBoundaries. We should reference the first comment of this thread into a new issue to track this when the PR is merged.