Add MRCA calculation based on Schieber-Vishkin algorithm

[jeromekelleher]

The Schieber-Vishkin algorithm allows us to compute the MRCA of two nodes u and v in constant time, given a O(n) preprocessing step and O(n) extra storage space. If someone is computing a lot of MRCAs, then it can be quite a bit speedup over just computing each one directly.

It would be nice to have an implementation of the mrca function that does something like this:

def mrca(ds, u, v):
    if "sv_tau" in ds:
          return _sv_mrca(ds, u, v)
    else:
          return _mrca(ds, u, v)

There's an implementation of the SV algorithm the tskit test suite here. This is a direct implementation of Knuth's version, which uses 1-based arrays instead of 0-based. I never got around to properly updating it to use 0-based, and just did a quick hack to get it converted. It would be nice to do this properly.

[benjeffery]

This is nice - and a great example of the extensibility of the dataset paradigm.