adjust_total_time before running on_complete callbacks

[pedropb]

Closes #83

What are you trying to accomplish?

We were using the on_complete callback to report the job total run time but we noticed that total_time only gets updated after the callbacks are called.

What approach did you choose and why?

There were 3 options:

1. Move adjust_total_time so that it's called before the callbacks.
2. Copy adjust_total_time so that it's called before the callbacks and also right before logging the output summary. This could be useful in case the "total_time after callbacks" is valuable.
3. Publish a new event via ActiveSupport::Notifications.instrument instead. This would overlap with the callback functionality.

I decided for 1 because I think it's the one that makes the most sense.

What reviewers should focus on?

• I'm not familiar with writing RubyGems so I added Timecop as a dependency without a group. If that's problematic I can try to come up with another solution to stub the passage of time.
• I would consider this a bugfix but if users are relying on the inaccurate total_time inside on_complete callbacks, this patch would be a breaking change.

[pedropb]

Latest pushed changes:

• Moved adjust_total_time from the completed? branch to the end of iterate_with_enumerator.
• Dropped TimeCop in favour of ActiveSupport::Testing::TimeHelpers.
• Consolidate the test files by merging the separate callbacks_test.rb logic into iteration_test.rb.

[GustavoCaso]

I would suggest simplifying the test:

def test_jobs_using_on_complete_have_accurate_total_time
    freeze_time do
      job = JobThatCompletesAfter3Seconds.new
      job.perform_now
      assert_equal(3, job.total_time)
    end
  end

That way you do not need to pass the assertion around, which is a little weird 😄

[pedropb]

This is a different test though and it would already pass on master without the patch.

The reason for passing the assertion around is so it can be injected in the on_complete handler to show that at that moment, it will/won't have the accurate total time.

[GustavoCaso]

Code LGTM. I would like it if you could modify the test file to remove the need to pass the assertions around. Apart from that this is ready to be merge 😄

[pedropb]

@djmortonShopify 2 questions before I hit merge:

1. should I bump the version in this PR?
2. what version bump (major, minor, patch) do you think this change should be?

[djmortonShopify]

@djmortonShopify 2 questions before I hit merge:

1. should I bump the version in this PR?

2. what version bump (major, minor, patch) do you think this change should be?

I think we usually bump the version in a separate PR. I'm ok with a patch version bump for this because I'm still of a mind that this is a bug fix. There are no changes to the API, so it'll only "break" for people relying on the previous, arguably incorrect, behaviour. @GustavoCaso Thoughts?