Pipeline.pipelines_by_branch() should use api.collect_all_pages()

Created by: mrj10

pipelines_by_branch() uses api.call(), which implicitly only returns the first page of GitLab API objects/resources (20 by default).

If a branch has more than 1 page worth of pipelines, `pipelines_by_branch() will return a subset, and if the SHA you're looking for isn't in the first page, you'll spin forever for a pipeline that is already there (just on a subsequent page).

I think calling api.collect_all_pages() instead of api.call() will fix it.

Probably worth scrubbing the rest of the codebase to use collect_all_pages() whenever pagination might be in play.