Clean up heron metrics for prometheus

[cckellogg]

Some metrics when exported to prometheus have the instance id in the metric name. This should be a label.

Examples: heron_client_stmgr_100_hello_messages_to_stmgrs heron_client_stmgr_1_hello_messages_to_stmgrs heron_execute_count_word_spout_default heron_execute_time_ns_word_default heron_fail_latency_word_default heron_execute_latency_word_default

[rohanag12]

As discussed in the meetup, it would be nice to clean up these metrics, some of them are really useful for performance tuning. Currently, it is not easy to visualize these metrics since prometheus does not support wildcards in metric names.

[rohanag12]

@huijunw @sreev