Add support for multi-instance-type clusters to AWS/GCP providers

[RobertLucian]

Benefits of the MITC functionality

1. Lower costs: Compared to single-instance-type clusters, reduces overall compute costs by using the most economical instance for the given workloads.
2. Simpler logistics: User no longer has to manage multiple clusters on their own.
3. Multi-purpose cluster: The cluster can now take any range of workloads. One cluster for everything (CPUs, GPUs, Inferentia). Just throw a bunch of node groups in the cluster config, and you’re set.

Example config

cluster_name: cortex
provider: aws
region: us-west-2

node_groups:
  - instance_type: t3.medium
    max_instances: 5
    min_instances: 200
  - instance_type: c5.xlarge
    min_instances: 5
    max_instances: 100
  - instance_type: g4dn.xlarge
    min_instances: 0
    max_instances: 100
    spot: true
  - instance_type: g4dn.xlarge
    min_instances: 0
    max_instances: 100
  - instance_type: inf1.xlarge
    min_instances: 0
    max_instances: 100

---

checklist:

• run make test and make lint
• update .circleci config
• test manually (i.e. test cluster(-gcp) commands, multiple node groups/pools, cpu/gpu/inferentia workloads in single cluster, etc)

[vishalbollu]

Awesome. Mostly nits. Can be merged after name field is set to required, memory map is tested.