Add additional metrics to ECS Integration

[reskin89]

Currently, statistics around the pull/start/stop time of containers is not present in the ECS integration. The datadog agent as a sidecar currently hits the metadata uri to get other statistics from the same endpoint.

I believe these additional metrics would be incredibly useful to determining slow starts/bottlenecks and the like, as I can see:

• how long my containers take to pull into ecs/fargate (since there's no true image caching, this matters far more than other platforms that cache container images per worker, its downloaded every time a task starts)
• how long before my container is considered started (good to narrow down boot time issues)
• how long before a container is stopped (such as seeing a cycling issue)