Convert the continuous integration process to use GitHub's internal
CI cloud.  Allows us to run up to 20 jobs in parallel, speeding up
CI immensely.

Translate all Travis CI configuration into event-driven GH YAML
files.

The original comments suggested adding a new SSH key to a user account
which would have allowed complete access to all repos and all other
global user settings.  A really bad thing if it got into the wild.

Instead, a single deploy key can be added to a single repo.  This only
grants git push access/etc. for that single repo and not for the entire
user account.  Update the docs accordingly.

Also rename push.yml->pull-request.yml since it runs on a pull request,
not a master push.

Add basic, fast TravisCI script to keep the TravisCI jobs alive even
if GitHub CI goes down.  Belt-and-braces, but it's free.

Limit the runs on the Travis jobs to 5 short ones in total to match
the runtime of the GitHubCI.

Testing with my local repo first

Testing now gets the proper release name for package(draft) and release.
Go back to the esp8266,github.io repo for release updates.