This prevented the -b option to install_stack.sh from working properly.

Now the install happens in the current directory, with logs in a 'logs'
subdirectory.

Other improvements: curl now fails if the Vagrantfile can't be
downloaded, because the RELEASE value is invalid.  Curl also doesn't
print as large a progress bar.  The script fails early if there's no
RELEASE specified.

(cherry picked from commit 66a25250)

Now the install happens in the current directory, with logs in a 'logs'
subdirectory.

Other improvements: curl now fails if the Vagrantfile can't be
downloaded, because the RELEASE value is invalid.  Curl also doesn't
print as large a progress bar.  The script fails early if there's no
RELEASE specified.

Now that it can use AWS tokens for auth, we need to be careful about
just dumping env into logs (and thus into splunk).

Because /var/home_jenkins on our jenkins test containers is mounted from
~/path/to/git/repo/.docker_volumes/jenkins_home you end up with an ssh
socket for ansible which is way past the 108 character unix limit for
filenames of sockets.

If you had a sandbox behind an ELB, this would take the whole sandbox
in/out of the ELB during an ansible run?

Also, this file/path won't exist later on sandboxes which also
source this file when running /edx/bin/update

(cherry picked from commit a8bd50f9)

Add docker_images script that absorbs some of the redundancy in balancecontainers and check_dockerfile_coverage.py