Various fixes to allows jenkins analytics to run analytics tasks on AWS resources

* Uses JENKINS_HOME=/edx/var/jenkins.  This addresses the pyinstrument
  re-install issue seen when JENKINS_HOME is a symlink.
* Installs libffi-dev and libpq-dev packages with analytics_jenkins playbook
* Adds an SSH Credential, which is passed as the MASTER_SSH_CREDENTIAL_ID
  parameter to the seed job.
* When the seed job is run, ignore unreferenced jobs, instead of deleting them.
  This allows you to run and re-run the seed job from Jenkins to update a
  particular analytics job, and untick the jobs you don't want to update.
* Lets <ANALYTICS_TASK>_EXTRA_VARS be a @path/to/file.yml