Add support for extended tasks information

[spapadop]

Currently, one can enable tasks collector with:

--collector.tasks
--tasks.actions="*" 

And the output metrics is like:

elasticsearch_task_stats_action{action="cluster:admin/snapshot/get"} 2
elasticsearch_task_stats_action{action="cluster:monitor/health"} 1
elasticsearch_task_stats_action{action="cluster:monitor/tasks/lists"} 1
elasticsearch_task_stats_action{action="cluster:monitor/tasks/lists[n]"} 3
elasticsearch_task_stats_action{action="indices:admin/mappings/get"} 1 

This is good enough to track how many tasks of each type is running at any time in your cluster, however it is not able to catch things like long running tasks. For example, imagine a long running search, indexing or reindexing task, that takes hours. One would like to be alerted whenever that happens so that an action is taken.

To achieve that, the labels that come with above metrics should be enhanced with information that are available on the GET _cat/tasks API, like task_id, type, start_time and running_time.