Revert introduction of forwardable

[scpike]

This PR attempts to fix an incompatibility between mongoid-locker >= 2.0.1 and mongoid-history. The issue was introduced by this PR. See mongoid/mongoid-history#238 (comment) for an example report.

Timeline

Here's a timeline of what I think happened:

1. Mongoid 7.1.0 introduced a regression (see https://jira.mongodb.org/browse/MONGOID-4849) by switching to the builtin ruby forwardable method of delegation rather than relying on the ActiveSupport-provided delegation used before.
2. mongoid-locker was made compatible with the regressed version of mongoid in Add forwardable module and upgrade rubocop #86
3. Mongoid then fixed their regression in 7.1.1+ with mongodb/mongoid@d078acd, which goes back to the active support way of delegating. (I'm not totally sure why mongoid-locker continued to work, but I think it's because requiring forwardable overrides what delegate does, so mongoid-locker gets its expected behavior but other gems have their delegate changed out from under them).

Underlying issue

The forwardable inclusion in mongoid-locker causes incompatibilities when you then include another gem that expects to be able to set up a delegator within a mongoid document class. For example the mongoid-history gem expects to be able to delegate using the active-support style here: https://github.com/mongoid/mongoid-history/blob/c8c4de1235c01252dfbdb9fe6c0abfc078ec1443/lib/mongoid/history/trackable.rb#L24-L25

Fix

This PR reverts the change to introduce forwardable to go back to using the ActiveSupport-provided delegate.

[dblock]

Thanks. Let's add a test that shows the problem?

[scpike]

Thanks. Let's add a test that shows the problem?

I couldn't think of a reasonable way to do this - can you point me in the direction you'd prefer? It seems strange and outside the scope of this gem to have a test that delegate :x, to: :y still passes after this gem is required.

I could add a test dependency on mongoid-history and try to set up a mongoid document that fails if forwardable is require'd. That's maybe a better integration test but also kind of strange.

[dblock]

Maybe just add a simple integration test ala https://github.com/ruby-grape/grape/tree/master/spec/integration?

[scpike]

This is required for defined?(Mongoid::History) to be true in the integration spec.

[scpike]

Maybe just add a simple integration test ala https://github.com/ruby-grape/grape/tree/master/spec/integration?

Ok I took a stab at it. If you cherry-pick 9cb3e2d over to master the new spec will fail.

[dblock]

Looking good! Fix CI, I have some nits and we're good to go, thanks for fixing this.

[dblock]

Is the if: needed? I expected for the test to be predictable.

[scpike]

I was setting this up to be similar to what you linked to over at https://github.com/ruby-grape/grape/tree/master/spec/integration, where you have a suite of integration tests where not all gems are required all the time, but rather required conditionally for the various combination tests.

It would still work to remove this check or even remove appraisal entirely.

[dblock]

Sorry for the delay, CI has an issue.

I have to approve it here in the upstream for first time contributors, but you can definitely see it in your fork when you push, so iterate to 🍏 . Thanks!

[dblock]

Merging, thanks, will deal with the jruby failure in #108

[dblock]

If you want to do some more work here, the test matrix could use an upgrade to include more recent versions of rubies/MongoDBs.