LBGPU: Don't forget boundaries on on_cell_strucutre_change()

[RudolfWeeber]

+Test

Fixes #2728

[fweik]

Shouldn't this call some interface function, instead of checking the lattice switch? Also how sure are you that no action should be taken for the GPU code in this case?

[RudolfWeeber]

Further offline discussion yielded the following:
While the GPU forgets boundaries when the event is triggered,
the CPU LB forgets the flow field.

The LBCPU cannot cope with changes of the parallelizaiton, and local and global box shapes.
It also needs to react to changes of the md skin due to the halo size.

The current event categories are too broad for this, so refactoring is needed, there. This also pertains to the Walberla LB integration.
Short-term, I'll introduce code to throw a runtime error, when the
event is triggered.

[RudolfWeeber]

This should go into 4.0.2

[codecov]

Codecov Report

Merging #2736 into python will decrease coverage by <1%.
The diff coverage is 100%.

@@          Coverage Diff           @@
##           python   #2736   +/-   ##
======================================
- Coverage      80%     80%   -1%     
======================================
  Files         508     508           
  Lines       27073   27076    +3     
======================================
- Hits        21822   21812   -10     
- Misses       5251    5264   +13

Impacted Files	Coverage Δ
src/core/event.cpp	95% <100%> (ø)	⬆️
src/core/grid_based_algorithms/halo.cpp	58% <0%> (-7%)	⬇️
src/core/grid_based_algorithms/lb_interface.cpp	74% <0%> (-1%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f2c8d12...751ac72. Read the comment docs.