visual regression testing

even if for now it screenshots the test index page and compares a PR against it. it'd be testing the reset mostly, so it'd be more like an e2e test, but that's coo until something stronger get's built.

Start by locating the test index page mentioned in the issue and determine how a screenshot comparison could run against a pull request. The initial scope is an end-to-end check focused on the reset; done means the comparison detects visual changes there, while leaving room for a stronger approach later.

Written by the indexing model from the issue text.