Command generation tests should assert against generated scripts rather than intermediate commands

[bartread]

Relates to findings from #12 and #13, where an existing command generation test - OneToOneRelationshipCommandGenerationTests.insert_with_fk_on_child_and_special_data_in_child_inserts_in_parent_and_updates_child - where test failed to catch problem described in #13.

Fundamentally I don't think asserting against operations and commands is robust enough. What matters is the output at the end of the process - i.e., the scripts that SimpleSave generates, and ultimately the resultant state of the database.

I think a good intermediate step for now is to begin transitioning tests to assert on scripts rather than commands. We can then later implement some localdb tests to check database state. (At the moment we're relying on PaymentSense internal tests for this, which we obviously can't use here.)