Report the collector as unhealthy if the number of running tasks is high

[fniessink]

Add an environment variable COLLECTOR_RUNNING_TASKS_HEALTHY_THRESHOLD and report the container as unhealthy if the number of running tasks (as tracked by Collector.running_tasks) exceeds the threshold.

Tasks:

• Add the environment variable to base_collectors.config
• Default value: 120. Rationale using the default config values: the collector starts max 30 tasks every 20 seconds that timeout after max 120 seconds so the theoretical maximum number of running tasks is (120/20) x 30 = 180. Use 2/3 of that. Add the rationale to the code and docs.
• Change the Collector.record_health method to not update the timestamp when the number of running tasks exceeds the threshold.