Use fewer q-Pochhammer calls when computing ghosts

[sflammia]

Currently a lot of time is spend computing finite q-Pochhammer symbols with large values of $n$. For non-principal ghosts, this makes calculation quite slow. We know that this can be avoided by using more compact formulas, but we need to derive these very carefully and then implement them.

[gskopp]

My gkshin branch is intended to address this issue. The function shin_of_tuple computes the values of the shin function needed for admissible tuples of reduced forms without needing any finite q-Pochhammer symbols. There is currently a bug in even dimension; Steve and I are both working on this.