Implement reverse-maximin ordering and corresponding Cholesky/linear KR-map estimate

[Blunde1]

Resolves #66

• Add notebook showing the various options for SPD precision estimation:
  • Direct and exact estimate. Ordering is AMD/METIS or similar from suitsparse
  • Incomplete Cholesky factor. Improves by expanding the graph
  • Approximate Cholesky factor using reverse-maximin ordering and corresponding sparsity structure
• Adopt reverse-maximin ordering into a separate module, cholesky_estimation
• Corresponding Cholesky estimation using the sparsity pattern and ordering.
  • First using linear KR-map methodology
  •  Secondly with Equation 2.3 in the reference paper by Schäfer
• Exchange reverse-maximin function in notebook with package/module implementation

Remember to document well. Reference the paper and Algorithm C.1 heavily