1. It's not just processes, it contains the list of all DMN Decisions as well.
2. It is closely linked to the SpecReference object that can be generated by looking through all the Spec files to find the processes and decisions they contain.
3. It is a cache of information, the file system is the source of truth. Seems likely we will cache more things in the future -- so setting things up this way made sense.
Removing (very nearly, except for script unit tests) all the XML Parsing we were doing, see related PR on SpiffWorkflow
Moved the Custom Parser into its own file to solve some circular import issues