Fix Bug in HPMG bottomsolve_gpu

[AlexanderSinn]

For some resolutions, this bug was causing out-of-bounds accesses, leading to wrong results or invalid memory crashes when running the CI tests on GPU. With these changes the CI passes when running on A100 with -DAMReX_ASSERTIONS=YES and -DAMReX_BOUND_CHECK=YES:

[asinn@max-mpag003]~/hipace% ctest --test-dir build --output-on-failure
Internal ctest changing into directory: /home/asinn/hipace/build
Test project /home/asinn/hipace/build
      Start  1: blowout_wake.2Rank
 1/20 Test  #1: blowout_wake.2Rank ..............................   Passed   17.29 sec
      Start  2: beam_evolution.1Rank
 2/20 Test  #2: beam_evolution.1Rank ............................   Passed    5.87 sec
      Start  3: adaptive_time_step.1Rank
 3/20 Test  #3: adaptive_time_step.1Rank ........................   Passed    6.38 sec
      Start  4: radiation_reaction.1Rank
 4/20 Test  #4: radiation_reaction.1Rank ........................   Passed    3.78 sec
      Start  5: grid_current.1Rank
 5/20 Test  #5: grid_current.1Rank ..............................   Passed    4.90 sec
      Start  6: linear_wake.normalized.1Rank
 6/20 Test  #6: linear_wake.normalized.1Rank ....................   Passed    5.58 sec
      Start  7: linear_wake.SI.1Rank
 7/20 Test  #7: linear_wake.SI.1Rank ............................   Passed    6.87 sec
      Start  8: gaussian_linear_wake.normalized.1Rank
 8/20 Test  #8: gaussian_linear_wake.normalized.1Rank ...........   Passed    5.65 sec
      Start  9: gaussian_linear_wake.SI.1Rank
 9/20 Test  #9: gaussian_linear_wake.SI.1Rank ...................   Passed    5.75 sec
      Start 10: reset.2Rank
10/20 Test #10: reset.2Rank .....................................   Passed    4.48 sec
      Start 11: beam_in_vacuum.SI.1Rank
11/20 Test #11: beam_in_vacuum.SI.1Rank .........................   Passed    5.82 sec
      Start 12: beam_in_vacuum.normalized.1Rank
12/20 Test #12: beam_in_vacuum.normalized.1Rank .................   Passed    6.63 sec
      Start 13: slice_IO.1Rank
13/20 Test #13: slice_IO.1Rank ..................................   Passed   12.47 sec
      Start 14: output_coarsening.2Rank
14/20 Test #14: output_coarsening.2Rank .........................   Passed    6.41 sec
      Start 15: beam_in_vacuum_open_boundary.normalized.1Rank
15/20 Test #15: beam_in_vacuum_open_boundary.normalized.1Rank ...   Passed    5.27 sec
      Start 16: laser_evolution.SI.2Rank
16/20 Test #16: laser_evolution.SI.2Rank ........................   Passed   11.74 sec
      Start 17: laser_ionization.1Rank
17/20 Test #17: laser_ionization.1Rank ..........................   Passed    6.13 sec
      Start 18: laser_STC.1Rank
18/20 Test #18: laser_STC.1Rank .................................   Passed   69.42 sec
      Start 19: plasma_density_from_file.SI.1Rank
19/20 Test #19: plasma_density_from_file.SI.1Rank ...............   Passed    6.79 sec
      Start 20: transverse_benchmark.1Rank
20/20 Test #20: transverse_benchmark.1Rank ......................   Passed   10.54 sec

100% tests passed, 0 tests failed out of 20

Total Test time (real) = 207.80 sec

• 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]

Thanks for the fix!