MRCA algorithm can be faster

[niemasd]

From what I understand from the mrca() function of a Tree object (https://github.com/jeetsukumaran/DendroPy/blob/master/src/dendropy/datamodel/treemodel.py#L3953), the current algorithm performs a top-down traversal of the tree and returns the last node it sees that has all user-specified taxa as descendants

I believe a faster algorithm (which is what I have implemented in TreeSwift) is the following:

for each node u:
    u.num_descendants = 0
for each leaf l:
    for each ancestor a of l:
        a.num_descendants += 1
        if a.num_descendants == total number of leaves:
            return a

[jeetsukumaran]

Thanks for this. Will look into this later!