Clarrify use of regression label

As a follow up from the discussions in https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/17530#note_83426018

This MR aims to clarify and simplify the use of the regression label.

We should not remove the regression label as part of the "Demotion" / "Rescheduling" process. If the issue was indeed a regression it should carry such context forward till it's resolved. The deliverable milestone is what's important here. The existence of regression just tells us that this ~bug broke something that was working before. The created at time of a regression is enough to tell us which release the regression was found in.

This would make the triage process of regression and ~bug the same. A ~bug with regression just tells us that something worked before and it needs extra attention from EM+PM to schedule/reschedule.