Fm/task/unst xxxx vectorize setumod

[FlorisBuwaldaDeltares]

What was done

• vectorized setumod (80% speedup!)
• Split of code path for dflowfm2d-noordzee in separate functions (icorio = 5, newcorio = 1, kmx = 0, jasfer3D=1)
• prefetch node-based arrays into link-based arrays for contiguous memory access
• no longer calculate coriolis if coriolis constant is 0
• speed up coordinate transformation by either hardcoding or using nod2linx/y_fast
• update few cases with extremely tight tolerances

Evidence of the work done

• Video/figures
  -
  dflowfm2d-noordzee_0_5nm-j22_6-v1a
  [ ] Clear from the issue description
• Not applicable

Tests

• Tests updated
  <add testcase numbers if applicable, Issue number>
• Not applicable

Documentation

• Documentation updated
  <add description of changes if applicable, Issue number>
• Not applicable

Issue link

https://issuetracker.deltares.nl/browse/UNST-9669

[thomas-pijls]

Nice work, you did a lot

I did mostly a style review; the exact workings of this code I am not familiar with. If you have any questions I am happy to discuss them with you

[V-Chavarrias]

Nice job. I have nothing major. Just some comments for me to learn. You can ignore them. I guess we have no testcase where we can check the impact on an analytical solution or something similar, isn't?

[FlorisBuwaldaDeltares]

Nice job. I have nothing major. Just some comments for me to learn. You can ignore them. I guess we have no testcase where we can check the impact on an analytical solution or something similar, isn't?

Thanks! no, there's a few validation cases which are quite sensitive but nothing analytical to unit test against. Would be nice in the future because this code is super duper fragile

[thomas-pijls]

Looks good to me 👍