Fix consistency issue between ActiveModel::Conversion and ActiveModel::Lint

[santib]

Summary

ActiveModel::Lint says #to_key should return nil if #persisted? is false, but ActiveModel::Conversion that implements #to_key doesn't care about #persisted? but it actually cares about #id.

The primary change in the PR is on ActiveModel::Lint#test_to_key to set the #id method on the model instead of the #persisted? method, so it always passes when ActiveModel::Conversion is included in a class.

[matthewd]

Relevant context for why it used to work that way is here:

rails/activemodel/lib/active_model/conversion.rb

Lines 43 to 44 in 6e6764b

	# Note the default implementation uses persisted? just because all objects
	# in Ruby 1.8.x responds to <tt>:id</tt>.

This seems like an improvement, though I am conscious of the fact that this could result in a model that previously passed the lint now failing it, because we're changing the implemented expectation to better match what the documentation claimed (and, almost incidentally, what our default implementation does). Hopefully that's okay in practice? 🤷🏻‍♂️

[santib]

Relevant context for why it used to work that way is here:

rails/activemodel/lib/active_model/conversion.rb

Lines 43 to 44 in 6e6764b

	# Note the default implementation uses persisted? just because all objects
	# in Ruby 1.8.x responds to <tt>:id</tt>.

Wow, very interesting, I wasn't aware of that.

This seems like an improvement, though I am conscious of the fact that this could result in a model that previously passed the lint now failing it, because we're changing the implemented expectation to better match what the documentation claimed (and, almost incidentally, what our default implementation does). Hopefully that's okay in practice? 🤷🏻‍♂️

hmm yeah, that's a valid concern. Other options could be:

1. Making ActiveModel::Conversion use persisted? in the to_key method
2. Changing the ActiveModel::Lint test to mock both persisted? and id to return false and nil respectively

I guess the safest one is 2. ?