Dict vs NamedTuple in DynamicACSetTransformations

DynamicACSetTransformations (Tight and Loose) currently use NamedTuples to store their components, but the spirit of Dynamic data structures is to not have data in the types, so these fields ought use `AbstractDict{Symbol}` instead. This requires some changes elsewhere in the code, such as `compose`, where small changes should be made to make code generic w/r/t Dicts vs NamedTuples (e.g. use `mapvalues(f,comp1,comp2)` rather than `map(f,comp1,comp2)`).