standardize selectors used for tests

We currently use semi-random selectors for accessing elements in tests. Having a documented standard way to access elements means fewer decisions, and make it more obvious when an element is accessed in tests.

from https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/17011#note_70472380

Imo, it would be even more ideal if we could do something like data-test attribute. I read about it in and https://docs.cypress.io/guides/references/best-practices.html#Selecting-Elements I think it's something we should consider