Separate host OS docker image from coala docker image

[jayvdb]

As the docker image is one complete script, each time a bug is fixed in coala or coala-bears for a release, the image is rebuilt and any new versions of the dependencies from openSUSE are automatically included in the next build.

e.g. #170 fixes 0.9, but merging that would result in all sorts of dependencies to also be updated, potentially breaking something.

This will be slightly improved by switching from Tumbleweed to Leap (#144) , but that reduces the pace of unnecessary change, without eliminating it. An OS release usually upgrades even major versions of software, and typically upgrades minor versions, whereas we really want only patch level upgrades to occur on our stable release images.

[jayvdb]

The easiest way to resolve this is to create a new docker repo which uses the docker image built in this repo at release time, and patches it slightly when there is a coala patch to the release, or a critical update is needed to a linter. i.e. minimising change and documenting each change that is made, rather than slurping in all sorts of changes from Tumbleweed (or Leap, for that matter) without being able to easily see what changed from the previous build of the same coala release.