major re-write on snaq check root

[crsl4]

We have a mess in the order of:

• check root
• merge leaves
• printing network

for the case of multiple alleles.
We need to really look into the order of things, and do not repeat steps.
Make sure that some things are done inside optTopRuns.

Problem: we need to print the network before merging leaves (which is currently done outside optTopRuns), so writeTopologyLevel1 needs to merge the leaves. But merging the leaves messes with the root.

Currently, we do:

• In optTopRuns:
  1. check the root
  2. print maxNet (with writeTopologyLevel1 which merges leaves and checks root again inside)
• In snaq:
  1. merge leaves for good before returning the network
  2. checks the root again

[crsl4]

From @cecileane comment:
This is unrelated to the docstring, but it would be good to include a formal test for the changes to writeTopologyLevel1 and to the end of snaq!, in 1da3325).

By the way, the chunk of code repeated in writeTopologyLevel1 and in snaq! could possibly be moved to checkRootPlace!: There would be less room for bugs (e.g. bug fixed in one place but not in the other, when the chunk is repeated), and the code for both writeTopologyLevel1 and in snaq! would be cleaner. The formal test could also be made cleaner that way.

[cecileane]

was this done via 1da3325 and #135?