Use MArrays for small data holders in marching_tetrahedra

[pjaap]

This reduces the wall time.

demo:

using ExtendableGrids, GridVisualizeTools

grid = uniform_refine(grid_unitcube(Tetrahedron3D), 6)

coords = grid[Coordinates]
cellnodes = grid[CellNodes]
func = ones(size(cellnodes, 2))

@benchmark marching_tetrahedra(coords, cellnodes, func, [[1.,1.,1.,-1.]], [])

yields

BenchmarkTools.Trial: 105 samples with 1 evaluation per sample.
 Range (min … max):  46.173 ms … 52.366 ms  ┊ GC (min … max): 0.00% … 2.72%
 Time  (median):     47.590 ms              ┊ GC (median):    0.00%
 Time  (mean ± σ):   47.930 ms ±  1.352 ms  ┊ GC (mean ± σ):  0.60% ± 1.22%

       ██▂▂▄▄       ▂▂▂▂ ▆                                     
  █▆▄████████▆█▆▆▆█▄████▄██▁▁▄▁▄▄▆▁▁▁▄▄▁▆▄▆▁▁▆▁▄▁▄▁▄▁▁▄▁▁▁▁▁▄ ▄
  46.2 ms         Histogram: frequency by time          52 ms <

 Memory estimate: 5.89 MiB, allocs estimate: 84.

without @MArray and

BenchmarkTools.Trial: 141 samples with 1 evaluation per sample.
 Range (min … max):  33.983 ms … 39.415 ms  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     35.359 ms              ┊ GC (median):    0.00%
 Time  (mean ± σ):   35.589 ms ±  1.086 ms  ┊ GC (mean ± σ):  0.82% ± 1.51%

       █▁▁▄▄▆  ▃▁▆     ▃     ▁▃    ▁                           
  ▆▆▆▆▆██████▇▇███▆▇▇▆▇█▆▇▆▄▇██▇▇▄▄█▇▄▇▆▄▁▁▄▆▁▁▄▄▁▆▄▆▄▁▁▁▁▄▁▄ ▄
  34 ms           Histogram: frequency by time        38.5 ms <

 Memory estimate: 5.89 MiB, allocs estimate: 80.

with @MArray.

This is reduces wall time by about 25%.

The cn array is not used, and therefore removed.

[j-fu]

Wow interesting didn't expect this, and for sure not this amount!

[j-fu]

... ok for me

[Copilot]

Pull Request Overview

This PR optimizes the marching_tetrahedra function by replacing regular arrays with static arrays (@MArray) for small, fixed-size data holders to improve performance. The change reduces wall time by approximately 25% while maintaining the same memory usage.

• Converts small fixed-size arrays to static arrays using @MArray macro
• Removes unused cn array variable
• Maintains same functionality with improved performance characteristics

[pjaap]

I also checked Copilot's comments. This does not speed up the method. So I will leave the @MArray macro. I think it is better readable.