more efficient scope chaining

[cammckinnon]

Currently, scope chaining is done by creating a new copy of the entire scope for each level of scope. That is, while it is O(1) to resolve symbols, it's O(n * k) to have n scope levels with k variables defined. I'm fairly sure that with a cleverer scope chaining datastructure, we could keep the O(1) symbol resolution but get O(k) for the scope levels. If not O(k) then definitely better than O(n * k) at least on average though.