Operational changes and perf updates

This had a lot of thrashing as we adapted it to run in our new infrastructure. Generally, we need:

• it to work on ECS with a Docker-based deployu
• to use resque-scheduler instead of Heroku scheduler (since the later doesn't exist on AWS :) (See #66, #63, #61, #59, #58, #57)
• Update dependent gems (see #60 and #68 )
• Reduce auto-refresh in the UI. On the failed page it no longer auto-refreshes, and on other pages, it does so only once per minute (see #70 )

Slight performance improvements

• When just getting counts, don't deserialize all the jobs (See #53)
• Various configuration things added to make it easier to use in Stitch Fix's infrastructure (see #51, #50, and #49). It's actually not straightforward how this should work as an open source Rails app, but the changes that might affect users are:
  • Change from Unicorn to Puma. Puma is both recommended by Heroku and works much better in our infrastructure
  • add a Dockerfile that is, sadly, not generically re-usable by anyone other than us. We're aware of that but doing so prevents us from derailing into a longer project.
  • Force SSL in the Rails code, which simplifies things for us
  • Migrate jobs we had configured in Heroku Scheduler to be in a resque-scheduler-powered configuration

v1.6.0: Merge pull request #43 from stitchfix/gem-update-2016-09

• Now uses the canonical Resque code that contains the merge of @davetron5000's PR (see #40)

v1.5.0: Merge pull request #35 from stitchfix/page-size-for-failed-queue

Adds ability to set page size on the failed queue page (See #35)

v1.4.2: Merge pull request #33 from stitchfix/bundle-update-2016-03

• Update dependencies to address Rails security issue

v1.4.1: Merge pull request #32 from stitchfix/monitor-sizes-directly

• Attempt to reduce memory footprint of monitoring jobs by using counts directly. See #32
• Disabled CI for Ruby 2.1 as travis is unable to build the JSON gem.

v1.4.0: Merge pull request #31 from stitchfix/failed-jobs-by-class

• Redesign of the internals of the monitoring to make it more clear how the monitors and notifiers should communicate (see #30)
• Can now stat failed jobs by class, not just as an entire count (See #31)

v1.3.1: Merge pull request #28 from stitchfix/security-update-2016-01-28-220800

Fixes the various Rails security issues that came up this week by updated gem dependencies.

v1.3.0

• Bash script to monitor queue sizes (see #20)
• Update dependencies (see #19)
• Make consistent with Stitch Fix OSS guidelines (see #21)

v1.2.0: Merge pull request #17 from stitchfix/feature/allow-alternate-url-format

• Allow new format for environment variables that's more compatible with Heroku's addon attachments feature: MYAPP_RESQUE_REDIS_URL is the preferred format.
• The old format (RESQUE_BRAIN_INSTANCES_myapp) still works.