Improve CI release process

[mrjones-plip]

Right now the process to do a new release is:

1. cut a new branch
2. add a new feature/bug fix to branch, including revving SemVer version in package.json
3. merge branch to master - this kicks off CI which publishes a new image to ECR
4. manually release a new version on GH which matches SemVer version in package.json

This has a few draw backs:

• Automation in step 3 to push a new image above doesn't detect if SemVer has been incremented, this will always publish an image, overwriting an existing image in ECR with a new version of the image re-using the existing version value.
• If you forget to do step 4 above, then the GH release version will be different than ECR

Suggestions to fix this Improve CI to:

• Check for existing SemVer in ECR and not push a new image if version already exists (also don't cut a new GH version per next item).
• Automatically create a new GH release when there's a new ECR release to ensure they're always in sync

[mrjones-plip]

We may choose to lock down PR formats with commitlint (see Watchdog example) which in turn could help us leverage Semantic Release (see watchdog example again)