Provide a function to convert mapping affine expressions to mapping filters

[ftynse]

#489 introduces more code that needs the mapping to be represented as both a mupa and a union_set, depending on the context of use. We now have (almost) duplicate code that traverses the tree to collect mapping information in different forms, and often call both. #518 changes the mapping node to no longer be a subclass of the filter node.

Provide a function to convert a mapping represented as mupa to a union_set form (inverse conversion is tricky) and remove duplication and extra tree traversals. Optionally, introduce an special "mapping" type that can be viewed as both mupa and union_set.