Support different serving images on a per-API basis

[RobertLucian]

Closes #900, #950.

This PR includes support for public-accessible docker images (i.e. DockerHub) and ECR images under the cluster's AWS credentials.

Here is a set of errors that could possibly be returned by this new feature:

# some common ones that could appear due to setting an invalid value for image or tf_serve_image keys error: cortex.yaml: api-1: predictor: image: xxxxx.dkr.ecr.eu-central-1.amazonaws.com/cortexlabs/python-serve:custom-d is not accessible error: cortex.yaml: api-1: predictor: image: xxxxx.dkr.ecr.eu-central-1.amazonaws.com/cortexlabs/python-serve:custom:::dsfds is not a valid docker image resource error: cortex.yaml: api-1: predictor: image: robertlucian/private-sample-image is not accessible # some errors that should never happen (mostly internal use) error: cortex.yaml: api-1: predictor: image: can't extract ECR credentials error: cortex.yaml: api-1: predictor: image: index out of range: can't extract ECR credentials error: cortex.yaml: api-1: predictor: image: illegal base64 data at input byte 2504: can't extract ECR credentials error: cortex.yaml: api-1: predictor: failed to encode docker login credentials: json: <some error> error: cortex.yaml: api-1: predictor: image: ServerException: failed to retrieve ECR auth token: <insert actual error> 

---

checklist:

• run make test and make lint
• test manually (i.e. build/push all images, restart operator, and re-deploy APIs)
• update docs and add any new files to summary.md (view in gitbook after merging)

[deliahu]

@RobertLucian wow, amazing job on this!! I can see the effort you put in to match our conventions and find all code that should be updated, thank you for your attention to detail!

I reviewed all but the docker image accessibility check; I'll check that out soon. The main comment is that Vishal and I spoke, and we think it probably makes sense to remove the "user-facing" images from cluster configuration now that we have it in the API configuration, what do you think? I'd be happy to discuss

[deliahu]

LGTM!

[deliahu]

@RobertLucian Thanks again for your great work on this!