Give TaskStep class knowledge of the underlying task

[agjohnson]

There is no self.retry on the task step class, this exists on the task.

This also fixes a bug where we weren't handling the lock on the python step as
well.

We discussed a better pattern to use here would be to rely on celery on_failure
and on_retry event methods more. This would simplify logic here greatly.

[humitos]

We discussed a better pattern to use here would be to rely on celery on_failure
and on_retry event methods more. This would simplify logic here greatly.

After taking a look at these changes. I think what I want is something like:

• keep the context manager do the work they are doing right now (handle BuildEnvironmentError, Warning, and thrown exception by us)
• use the on_failure to catch all unhandled exceptions plus all the exceptions (known) that we are handling in the task itself (e.g. LockTimeout and others)

[agjohnson]

Yup, agreed. I think the first pass at this solves the email notification issue as well as some general code cleanup around exceptions