Increase graceful shutdown timeout for prometheus/prometheus2 jobs

[ybykov-a9s]

I noticed that in some cases 25s given to prometheus to shut down using any kind of bosh stop might be not enough. In this case further logic runs kill -SIGKILL against prometheus process. It also has two downsides:

• 25s + time used by monit and time used by bosh-agent together might overlap default bosh-agent timeout 30s, and bosh task fails in that case.
• killing prometheus process with -SIGKILL leads to data loss on big deployments, because prometheus was not able to flush all WAL chunks from memory to the disk.
• hard killing prometheus using -SIGKILL most likely would force prometheus to rebuild its WAL from scratch on the next boot.

There is no strict recommendations how much time prometheus might need for a proper graceful shutdown, especially in big environments.

• Official Core behaviour: Prometheus does not have an internal maximum shutdown timer flag, it will continue to run until it finishes flushing memory blocks and WAL segments to disk
• Prometheus Kubernetes Operator: PrometheusSpec defines default value for terminationGracePeriodSeconds to 600s.
• Prometheus community kubernetes helm chart has the value of terminationGracePeriodSettings for the server set to 300s.

This PR increases time for a graceful shutdown from 25s to 110s. It also configures bosh-agent to wait 10s more than graceful shutdown is configured before reporting that the job failed. Current leftover 5s length is not always enough for bosh-agent. Updating it to 10s should be enough to never give up in most cases.

The PR also makes the graceful shutdown timeout configurable.