Try to split into three separate components -> controller, tester,
and 'tasks' (postgresql-tasks in our case).  The controller
component is the main part which is able to run the task remotely.
Tester is more-like library for 'tasks' component (should be
reusable on the raw git level).

* controller: Almost separated component.
* postgresql-tasks: Likewise.
* tester: Likewise.

* runner/result_templates/html.tmpl: When particular task results
are not available, print NOT AVAILABLE instead of FAIL.

Controller script runs the script on remote machine (OpenStack),
downloads the results, stores the result into its own
result-database and re-generates statistics for runs done so far.
It will be able to upload the results to "presenter" machine.

* config.sh.template: New doc file.
* controller: New file (the central script for CI).
* runner/result_stats: New file. Based on downloaded results from
testing machine, it generates single html file with stats.
* runner/result_templates/html.tmpl: New file. Template for ^^^.