Automate and improve our release process

At this point, only one maintainer is able to create a release for kops. This is a risk that needs to be resolved. Risks include:

- single point of failure
- if we have a p0 bug only one person can release
- if the user wins the lottery we have no release avenue

The release includes both the primary ami that is used for kops, and kops itself. The primary ami is built using tooling out of https://github.com/kubernetes/kube-deploy/tree/master/imagebuilder

Also, the kernel for the ami is not mainstream, but a custom build.

## Proposed Solutions

### Interm

1. we need to document existing process
2. determine which shared file repository and container registry we can use for the release
3. We can automate a release process via Travis, but it seems kinda complicated.

_Any other ideas_

### Long Term

We have multiple smaller projects within k8s now that will all need a process. We can possibly start with looking at what mini-kube and helm are using. I believe this should be under the purview of sig-release.

_Really need ideas for this_

No repository file, test, or entry point is named. Start by documenting the existing release process, then inspect the kube-deploy imagebuilder tooling referenced in the issue and compare release approaches used by minikube and Helm. The issue would be done only after an agreed, repeatable process covers both the primary AMI and kops releases.

Written by the indexing model from the issue text.