Extra arguments to map_ancestors to control behaviour of marginal cases

The current implementation of map_ancestors has the following behaviour for combinations of nodes from samples and ancestors:

Type of relationship	Shown in output of map_ancestors
ancestor->sample	Always
ancestor1->ancestor2	Only if ancestor2 has a relevant descendant
sample1->sample2	Always
sample->ancestor	Only if ancestor has a relevant descendant

(I designed the algorithm with a situation in mind in which only the first row is relevant: where all samples are taken from a single time point, and all ancestors are taken from an earlier single time point.)

However, it shouldn’t be too difficult to specify different behaviours under these other cases with extra inputs to map_ancestors - I’m happy to look into this later on if this seems useful.

[petrelharp]

From https://github.com/tskit-dev/tskit/issues/262

Nope, the function will fail if it is passed no ancestors. If you want to find relationships between all samples, you should just pass samples and ancestors the same list of nodes.

Oh! Let’s add this note to the documentation! This is very useful.

[petrelharp]

This table is extremely helpful; good idea. I’m not sure it’s correct, though? I will comment in the code.