Finalise API for "skipping" on Tree.seek()

[jeromekelleher]

As discussed in #2911 we have the skip option now in Tree.seek (and seek_index) which will use a more efficient algorithm when doing long-range seeking. We haven't really characterised how much more efficient it is and whether it's less efficient in any given situation. There's also the potential footgun of edges being added and removed in non-time sorted order to navigate.

So, my take is that we leave the default as using the linear algorithm with the option to skip. To close this issue we need to

• Do some basic performance benchmarking (how long to do short, medium and long-range seeking from a tree in the middle of the sequence). These can be done with (say) seek_index(index +/- j) for j in [1, 2, 10, 100]`` for a tree with (say) 1000 samples.
• Document the option, making the sortedness changes clear

We should add this to the next release milestone now as #2911 has been merged