Follows up from CRW-879

Today the che-theia-sources job builds three sets of assets for theia-dev, theia, and theia-endpoint.

If one fails, the build fails, and the whole process (which can tave over 9hrs) needs to start over at the beginning.

This is of course painful and resource-locking for no good value.

Splitting the job so that we can trigger dev, theia, or endpoint (or the last 2, or all three) would permit us to re-run failing parts faster.

Requirements:

• DONE: build.sh now publishes its intermediate images to quay so the next stage can use them outside of a jenkins job node context
  • eg., https://quay.io/crw/theia-dev-rhel8:2.9-pr-builder-x86_64 or https://quay.io/crw/theia-dev-rhel8:2.9-tmp-builder-x86_64
• DONE: new script to collect assets from containers - see https://github.com/redhat-developer/codeready-workspaces-theia/blob/crw-2-rhel-8/collect-assets.sh

• DONE: new flag to build.sh – docker rmi from local cache before building images - --rm-cache added
  • leave system prune in jenkinsfile - sh "docker system prune -af || true" (maybe not needed?)
• DONE: Jenkins string param PLATFORMS: arches to build (x/z/p)
• DONE: Jenkins string param CONTAINERS: images to build (theia-dev, theia, theia-endpoint)
• DONE: do not regen container.yaml based on arches flag

Future items moved to CRW-1905