ia() and get_parents()

[nthiad]

This patch adds ia(term_type, graph) and get_parents(term_type, graph) for #28 based on https://www.kaggle.com/competitions/cafa-5-protein-function-prediction/discussion/405237 (as best as i could interpret this).

Here's what it looks like to use:

>>> from manas_cafa5.protein import Protein; p = Protein.from_file('p68510', 'data/p68510.xml'); g = Protein.build_graph('data/go-basic.obo')
>>> p.ia('go',g)
7.111309583204707

and this feature can be used for the analysis requested in #28:

>>> len(p.go_terms())
29
>>> p.ia('go',g) * len(p.go_terms())
206.2279779129365

This patch doesn't implement the jupyter notebook but should make that notebook easier and more clear to compose.

[leandroradusky]

Here we do not want the actual terms: we will use this function in order to compute candidate terms, the actual terms for a protein are not candidates, they are already assigned to the protein then they will give us no score in the predictions. We just want the ancestors.

Also, will be good to have a parameter distance like the descendants function where we compute the ancestors up to any specific degree.

[leandroradusky]

Just one comment and then we're good to 🚀 :)