Implementing MARK03 Experiment

[gordonkoehn]

This PR implements/closes:

• MARK03 - Better Initialisation Strategies #75

will also address

• Run MARK03 #99
• Test MARK03 #98

[gordonkoehn]

As of now, MARK03 is implemented with RANDOM, DEEP, STAR and HUNTRESS trees - the MCMC5 trees are still completely missing.

There are some issues with the huntress trees. Apparently, HUNTRESS does not guarantee to include all mutations in the mutation matrix. This brings additional complexity. To nevertheless start an MCMC in those cases, one would have to remove this mutation from the mutation matrix - otherwise, that is not compatible with SCITE. This appears cumbersome and flawed in many ways.

Hence, I propose to catch those cases and plot the final log_prob / metric to iteration plot without the Huntress tree in those cases.
Therefore, it appears best to run the mark03 experiment always with snakemake --keep-going mark03 to allow the failure of some process branches, whilst also implementing the plotting of the final plots without huntress once per default.

Also next

• document and comment well
• remove unnecessary HUNTRESS ID from _file_id.
• add the MCMC5 tree generator
• fix position of legend

[gordonkoehn]

Came across:

• Refactor tree_from_tree_node to be static of Tree #65
  is resolved here.

[gordonkoehn]

yg.tree_inference.evolve_tree_mcmc() implements the mcmc tree evolver.

Need to

• integrate into the MARK03 workflow now.

[gordonkoehn]

This PR also removes snakefmt again / deactivates it for now as the code is intended where it should not be. Issues have been raised with developers:

• Intendeation in run directive breaking code - simple declarations snakemake/snakefmt#197