Fix behavior of _json_schema_to_grammar for schemas with many properties

[hudson-ai]

The conditional for adding commas to incomplete sets of properties in _json_schema_to_grammar._process_object was incorrect, leading to missing commas in the grammar when the number of properties was too high.

Added a test that fails on the current main branch.

@riedgar-ms please take a look when you can

[riedgar-ms]

That is embarrassing.... thanks for catching this. While you're at it could you also add an explicit test for an object with a single property, please? It looks like some of the other tests do have such a thing, but an explicit test case would be good to have.

[hudson-ai]

@riedgar-ms done. Let me know if you think it's worth parametrizing these two tests with different "type" fields or if you think that's well-covered by other tests.

And not embarrassing at all... I read that bit of code pretty closely myself and wouldn't have caught the problem without dogfooding this corner of the library.

[riedgar-ms]

Thank you. I think that this is adequate for now - we do have tests working with nested types and the like, so unless we have a failing case, testing effort is probably better spent elsewhere.