Respect mentor preference for single/double pairing

[kfanning]

Our survey asks mentors if they are ok with getting two mentees (since usually we have more mentees than mentors, some double pairs are needed). Currently the parser and algorithm ignore this preference, requiring some manual tuning of the out put as was done this year.

It should be easy to respect this response and only double pair mentors who are ok with it. Just want to do it in a clean way.

Similarly maybe could allow for n-tuple pairs instead of hardcoding double max? Unlikely needed but might look nicer.