Should we instrument the reason for interruption?

[sambostock]

As per #240 (comment)

As a user of the library, it would be helpful if there was an easy way to get notified (as in ActiveSupport::Notifications.instrument) when a job is interrupted due to elapsing the max run time setting. For example, a user could track how many times this is happening per some application-level metric to tune the time out setting accordingly.

We already instrument interruption, but we do not capture the interruption reason. We could potentially do this in job_should_exit? and forward the value to the instrumentation. We could also consider having the interruption adapter return a string or symbol, rather than just a boolean.