Separate Clade and Constraint

[bredelings]

Currently the Clade object does double-duty, serving as both a clade and a constraint. We should separate these two roles. Many of the problems that caused #286 come from treating all constraints as if they were a clade, when they actually are not.

Currently a Clade object can be a normal clade, or it can be a negative constraint (NOT X), or it can be an "optional" constraint (A OR B OR C). We should separate Clades from Constraints by making new classes for constraints:

• CladeConstraint(Clade)
• NegativeConstraint(Constraint)
• OrConstraint(vector)

We should also separate Clades and Constraints at the Rev language level, instead of using just the function clade( ) to construct all of them.

A lot of things that make sense for clades don't make sense for Constraints. For example, the taxon set for a NegativeConstraint doesn't really make sense.

Also, clades are normally considered to be the rooted form of a bi-partition, separating the include group (the clade taxa) from the exclude group (the other taxa) and the root. A clade should perhaps specify both the include group and the exclude group.

See some of the problems mentioned in #215 as well.