Avoid duplication of combinatorics for charge-symmetric final states in Hlt2 combiners

And if possible, find a way to do so without applying decay descriptors that are non-physical or that do not represent the intended physics content of the reconstructed decay.

Excerpts from discussion on !218 (merged)

@spradlin comment: A minor comment that everyone who has ever tried to write a charm line has received: when your final state is charge symmetric, like K- K+ or K- K+ pi- pi+, using charge-conjugate decay descriptors with 'cc' will duplicate the combinatorics. This can be seen to happen in lines 852 and 853. We usually work around this by constructing, e.g., only the combinations labeled as D0 and using unphysical descriptors like 'D*(2010)- -> D0 pi-' in subsequent steps. I rather hope there will be better ways to avoid duplicating combinatorics in the future.

and

@apearce comment: .... I would really like to have a more semantic way of expressing this relationship ("this final state cannot determine the flavour the parent, but the D^{*} decay resolves the ambiguity"), maybe this is already possible with some Python wrappers so it at least looks like what we want, but I'm sceptical.