Fix possible NRE in expandable field accessors

[ob-stripe]

r? @remi-stripe
cc @stripe/api-libraries

A few changes to avoid possible NullReferenceExceptions when using accessors for expandable fields:

• add safe navigation operator to all getters
• use new helper methods for all setters:
  • when setting the ID, SetExpandableFieldId will create the ExpandableField<T> instance if it doesn't exist already, and resets the expanded object if the IDs don't match
  • when setting the expanded object, SetExpandableFieldObject will create the ExpandableField<T> instance if it doesn't exist already.

I've written some generic tests for StripeEntity to test the helper methods, and also a few tests for Customer.DefaultSource so that we test at least one "real" expandable field. It doesn't seem realistic to write tests for every single expandable field accessor, so I guess we'll just have to be careful with copy/paste when adding new expandable fields.

(I have a vague idea for simplifying the expandable field accessor mess, but I'll keep that for a future major version.)

(Also, I think the entity classes should not have public setters at all and should be implemented as immutable value objects, but that would require writing very long constructor methods, so not going to do that until we can codegen .NET.)

Fixes #1659.

[remi-stripe]

This looks good to me and thanks for adding the detailed tests.

OOC is it possible to ask one of the affected developers to test your fix before we release to make sure it covers their use-case?

[ob-stripe]

Released as 27.1.3.