Deprecating treeherder's artifact API

[William Lachance]

Hey all,

Following from bug 1258861 (stop storing text log artifacts) I'm
deprecating the main remaining uses of treeherder's artifact API (bug
suggestions and "build api" artifacts). I believe no one was actually
using these, but just in case:

* The buildapi artifacts stored the buildbot "buildername" and request
id (for retriggering). The buildername was already stored in the per-job
ref_data_name, so the only important bit was the request id (which is
now stored in a job detail called buildbot_request_id). Here's an
example of retrieving this information for a particular job:

https://treeherder.allizom.org/api/jobdetail/?job_guid=176d846fbec7729736fc31c61baf2ef2c5bf5747&title=buildbot_request_id

As mentioned, you can get the ref_data_name from the job itself. Here's
an example:

https://treeherder.allizom.org/api/project/mozilla-inbound/jobs/32839881/

For more information see:
https://bugzilla.mozilla.org/show_bug.cgi?id=1301729

* The bug suggestions artifacts will no longer be generated. Instead, we
will generate them on-demand, caching as needed so retrieving them will
be fast. You will get them at a "bug suggestions" endpoint, like this:

https://treeherder.allizom.org/api/project/mozilla-inbound/jobs/32839881/bug_suggestions/

For more information see:
https://bugzilla.mozilla.org/show_bug.cgi?id=1281820

--

In general, storing large "artifacts" is not a good fit for something
like Treeherder, which is oriented towards displaying jobs and
classifying them. There are other services that are much better at
storing generic types of data.

If you have a small piece of data (or a URL) you want to associate with
a job, use a "job detail" (you can store this by submitting a "Job Info"
artifact:
http://treeherder.readthedocs.io/submitting_data.html?highlight=job%20info#job-artifacts-format).
If you have a large piece of data you want to store, use something like
Amazon S3 (and link to it from a job detail).

For now at least, the artifacts API will still be used to *submit* job
metadata, but the information will now be stored in specialized database
tables.

Will