Add more scheduler metrics.

[xujyan]

• Refactor the metrics variables to use a Metrics object nested in the scheduler to hold them.
• Add some essential scheduler metrics such as uptime and task lost/failed/finished, etc.

[xujyan]

The metrics look like this now:

{
scheduler.total_requested_mem_mb: 700,
scheduler.tasks_killed: 0,
scheduler.tasks_launched: 1,
scheduler.tasks_lost: 0,
scheduler.offers_unused: 385,
scheduler.uptime: 1944.163665,
scheduler.total_requested_cpus: 1,
scheduler.cluster_count: 1,
scheduler.total_requested_disk_mb: 3072,
scheduler.framework_registered: 1,
scheduler.tasks_failed: 0,
scheduler.tasks_finished: 0,
scheduler.resource_offers: 386,
scheduler.offers_incompatible_role: 0
}

[xujyan]

@vinodkone this is a followup to #32, could you take a look?

[vinodkone]

:+1: