Treeherder prod pushes 13th Oct -> 12th Nov

[Ed Morley]

The Treeherder team had a work week last week (hence the delay of this
email), where progress was made on the auto-classification project.
Prototypes to follow soon!

User-facing changes from recent deploys:
* Job group collapsing now avoids collapsing retriggered jobs, so they're
harder to miss.
* The messaging in the error summary panel when logs are missing is now
correct (previously said panel would refresh).
* Two people submitting the same bug classification at the same time no
longer gives an error.
* API: Performance of the /jobs/ endpoint has been improved, which should
help with UI-generated queries too.
* API: The max request time has been lowered from 120s to 30s to reduce
impact of both accidental and intentional API abuse.
* API: Support for Hawk authentication has been added (using OAuth
credentials is now deprecated and will be removed imminently).
* API: Server exceptions generated when interacting with the API now
correctly appear in New Relic, so are easier to diagnose. If you are a
submitter, we can add you to the New Relic account to ease debugging.
* API: The /resultsets/ endpoint now supports both 12 and 40 character
revision SHAs.
* API: Support has been added for ingestion of jobs from a Pulse stream.
* Perfherder: Many more fixes than I can list here, see write-ups for the
big items on Will's blog: http://wrla.ch/blog/category/perfherder/

Full changelog (220 commits; lots of backend/infra improvements not
mentioned above):
https://github.com/mozilla/treeherder/compare/656a91e...c27b3e8

If you see any issues, please file bugs in Tree Management::Treeherder*

Many thanks,

Ed