Composite morphisms and attributes in C-set accessors

[epatters]

Resolves #294. See the unit tests for examples.

[jpfairbanks]

Looks great! For the incident(acs, part, [:f, :g]) implementation, we should add a note about the order of composition being incident(acs, incident(acs, part, :f), :g) which is like inv(f) \cdot inv(g) which people might expect to be backwards. I wasn't sure about the order because chaining incident is less intuitive than chaining subpart.

[epatters]

I went back and forth on the order for incident but ultimately decided that the least confusing thing was to make the order [:f,:g] consistent with compose(f,g). I'll add a note to the docs about this.