IDL code generation has been a source of minor bugs and is perhaps not as well-covered by unit tests as it should be. In idl/tests/test_generator.py, at a minimum, we should add:

• Test generated parse and serialization functions for relevant combinations of IDL field types (e.g. variant<array<>>)
• Testing for constructors which use both NamespaceString and DBName parameters
• Test generated code for built in IDL validators at edge cases
• Test IDL files directly from a path
• Test the generated code for the definitions of the IDL structs found in unittest.idl