Deterministic test for bicoloring

[gdalle]

We should add a small deterministic test for bicoloring, like @amontoison's rectangle. The issue is that for NaturalOrder and LargestFirst I get trivial colorings. Which order did you consider to make the nice pictures?

[gdalle]

Here's a playground if you wanna try it:

using SparseArrays, SparseMatrixColorings, LinearAlgebra, Test

problem = ColoringProblem(; structure=:nonsymmetric, partition=:bidirectional)

A = spzeros(Bool, 100, 200);
A[1, :] .= true;
A[2, :] .= true;
A[end-1, :] .= true;
A[end, :] .= true;
A[:, 1] .= true;
A[:, 2] .= true;
A[:, end-1] .= true;
A[:, end] .= true;
display(A)

algo = GreedyColoringAlgorithm(RandomOrder(); decompression=:direct);
result = coloring(A, problem, algo);
maximum(column_colors(result)) + maximum(row_colors(result))  # always << 100

algo = GreedyColoringAlgorithm(NaturalOrder(); decompression=:direct);
result = coloring(A, problem, algo);
maximum(column_colors(result)) + maximum(row_colors(result))  # 101

algo = GreedyColoringAlgorithm(LargestFirst(); decompression=:direct);
result = coloring(A, problem, algo);
maximum(column_colors(result)) + maximum(row_colors(result))  # 201