Implement load balancing mechanism

[cgiovando]

This is particularly important if implementing a more dynamic solution, where processing resources need to scale to support demand efficiently. Something that can be done with Elastic Load Balancing for the AWS-hosted OAM, and possibly with multithread strategy when done on a local computer.

[mojodna]

ELB and HAProxy are straightforward ways to implement this (recommended for both hybrid and dynamic).

Implementing appropriate auto-scaling rules (or finding ways to use Lambda either for all traffic or as a supplement) is going to be the primary challenge here:

• what criteria can be used to scale up quickly enough to respond to spikes?
• what criteria should be used to scale down?

[hectcastro]

The API servers will have processes to publish CloudWatch metrics on SWF decider and activity worker task list counts. The plan is to reuse those for scaling the ASGs for deciders and activity workers.

See also: https://github.com/hotosm/oam-server-deployment/pull/1