Improve project testing and automations

Description

Running on the runner instead of in a container (see also https://github.com/solvaholic/octodns-sync/issues/50#issuecomment-830710533, #53) drastically changes what a "release" is for this project, as well as what tests and automations are required.

Expected Behavior

Meaningful tests run for any changes pushed to this repository; Where appropriate these serve as required status checks in pull requests.

The most recent numbered release can be specified in workflows by various labels. For example: @2.2.0, ~@2.2,~ and @2 are synonymous when v2.2.0 is the most recent release.

The linter and other frequently-run tests are fast.

Actual Behavior

Currently (04 May 2021) the functions of this action are not tested automatically, the release workflow is wrong, and the linter is slow af.

Possible Fix

Automate integration and validation testing; Implement as required status checks in pull requests.

Automate release tagging.

Find a faster way to run super-linter, or run individual linters on the runner.

Update Makefile to run those :point_up: locally.

Checklist

[x] Lint faster: Are shellcheck and yamllint all we need?
[x] (#33) Automate syncing major and minor version tags when a release is published or unpublished
[x] Document manual procedures, including rollbacks
[ ] Ensure automated actions are resilient and can't break much when they run amok
[ ] How much of octodns-sync can be tested without creds and a DNS provider?
[ ] Add validation testing on .. all pushes? At least on pushes to PRs and to main
[ ] Organize workflow triggers and required status checks
[ ] Gravy: Report performance of test runs, and correlate changes with changes

solvaholic / octodns-sync