Add NearestNeighborCluster algorithm

[mkborregaard]

This is the final landscape algorithm. Maybe some testing to work out whether it really gives the same results would be in order. One difference in behaviour is that the original code applies the mask first here. That means that for masks with many NaNs, this code here might end with numbers of cluster that for reasons of random sampling are more spread out around the target value of p. Ameliorating this would require making a specific rand! method for this particular NeutralLandscapeMaker, so it loses some consistency. Thoughts?

plot(
   heatmap(rand(NearestNeighborCluster(0.4), 50, 50)), 
   heatmap(rand(NearestNeighborCluster(0.5), 50, 50)), 
   heatmap(rand(NearestNeighborCluster(0.6), 50, 50)), 
   layout = (1,3), size = (1200, 400), colorbar = false, title = ["p = 0.4" "p = 0.5" "p = 0.6"]
)

[codecov-io]

Codecov Report

Merging #41 (41216e9) into main (d901551) will decrease coverage by 0.47%. The diff coverage is 0.00%.

@@           Coverage Diff            @@
##            main     #41      +/-   ##
========================================
- Coverage   7.32%   6.84%   -0.48%     
========================================
  Files         12      13       +1     
  Lines        273     292      +19     
========================================
  Hits          20      20              
- Misses       253     272      +19     

Flag	Coverage Δ
unittests	6.84% <0.00%> (-0.48%)	:arrow_down:

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
src/NeutralLandscapes.jl	100.00% <ø> (ø)
src/algorithms/nncluster.jl	0.00% <0.00%> (ø)
src/algorithms/nnelement.jl	0.00% <0.00%> (ø)

---

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 d901551...41216e9. Read the comment docs.

[gottacatchenall]

Maybe some testing to work out whether it really gives the same results would be in order

can this be done with PyCall? seems like a way to call NLMpy/NLMR equivalents would be good for testing and benchmarking

[mkborregaard]

It's always problematic when random number draws are involved I think - unless we cache a few million random numbers and read from that instaed of rand calls