If no pipeline hooks are active for a project, we still go through the
trouble of generating the hook data, which may make a number of SQL,
Gitaly, and Redis calls unnecessarily.

Most projects don't have any hooks, so we can reduce the amount of load
by only building the hook data if there are active hooks or
services. This requires several cheap SQL queries. Note that the optimal
way would be to lazily load the hook data, but that requires more
complex changes.

This is a similar optimization that was implemented for PostReceive
(https://gitlab.com/gitlab-org/gitlab-foss/-/merge_requests/31741).

Relates to
https://gitlab.com/gitlab-com/gl-infra/infrastructure/-/issues/12055