Revisit pickling logic for TraitListObject

[mdickinson]

This is a reminder to myself to revisit the pickling infrastructure for TraitListObject at some point before the 6.1.0 release. In particular:

• we don't have good tests for pickling and unpickling; there are some indirect tests, but it would be good to have direct unit tests in test_trait_list_object
• with the right unpickling setup, it shouldn't be necessary to verify that self.trait exists when validating the length
• I'd like to understand better the need for the object and trait manipulations in __getstate__ and __setstate__.