Add provisioning retries to beaker plugin to overcome aborted beaker jobs

[happz]

As discussed with several beaker plugin users, a single failed Beaker job means an end of the provisioning, even though a retried, new job might succeed. tmt lacks beaker-jobwatch capability to deal with unsuccessful reservations:

• submit new job when current one aborts
• there must be a limit on retries, controlled by a key + sane default
• if possible, new jobs should avoid machines on which the aborted jobs crashed - might be optional, controlled by a flag-like key
• it probably makes less sense to retry a job that spent its lifetime in queue and never run
• nice logging of job details and task results would be nice, if only within the debug/verbose mode