BpmnParser.find_all_specs() breaks with nested subprocess

[asauve2]

During the loading process of a workflow a dictionary is updated while
being iterated in BpmnParser.py

This results in

• RuntimeError: dictionary changed size during iteration

when there are two levels of nestings of subprocess workflows.

As I cannot post pull request, the fix consist only in making a copy of dict keys
that would look like the snippet below:

    def find_all_specs(self):
        # This is a little convoluted, but we might add more processes as we generate
        # the dictionary if something refers to another subprocess that we haven't seen.
        processes = dict((id, self.get_spec(id)) for id in self.get_process_ids())
        while sorted(processes.keys()) != sorted(self.process_parsers.keys()):
            for process_id in tuple(self.process_parsers.keys()):   # <===============
                processes[process_id] = self.get_spec(process_id)
        return processes

As I'm not sure if the python API would return consistently the keys in the same order
I also added a sorted() call for the comparison. Any comment on this part would be welcome.

[essweine]

I've incorporated your proposed fix in f113aba

Someday, we'll replace this parser with something a little easier to deal with.