Fully constructed pipelines for all release-tasks

The release tasks are split into several pipelines. Partly this is due to certain tasks taking place at different times - maintenance branches are created at feature complete, while the rest of the release process happens later.

The full release pipeline is initially split in two to allow a human to function as a 'circuit breaker' and for extended validation of the release candidate. The first part simply creates the release branch and updates the maintenance branches for the next release. The second part does the bulk of tagging, building, and shipping the agent. These pipelines should be merged in the future when we no longer require that level of human involvement.