fix(owners): Remove repo separation and fix unreleased commits

[dcramer]

• Refactors various fixtures to be more dynamic/flexible
• Removes commits endpoint
• Only identifies commits from repos with any releases

[dcramer]

[dcramer]

EXPLAIN (ANALYZE, BUFFERS) select c1.*
from sentry_commit c1
join (
    select max(c2.date_added) as date_added, c2.repository_id
    from sentry_commit as c2
    join (
        select distinct commit_id from sentry_releasecommit
        where organization_id = 1
    ) as rc2
    on c2.id = rc2.commit_id
    group by c2.repository_id
) as cmax
on c1.repository_id = cmax.repository_id
where c1.date_added > cmax.date_added
and c1.author_id IN (
    select id
    from sentry_commitauthor
    where organization_id = 1
    and lower(email) IN ('dcramer@gmail.com', 'david@sentry.io')
)
order by c1.date_added desc

[dcramer]

follow up will be to write a cache for "Latest Released Commit" to avoid this query:

    select max(c2.date_added) as date_added, c2.repository_id
    from sentry_commit as c2
    join (
        select distinct commit_id from sentry_releasecommit
        where organization_id = 1
    ) as rc2
    on c2.id = rc2.commit_id
    group by c2.repository_id

[mattrobenolt]

It might be worth noting that Django's default behavior of iexists and friends is to use UPPER() instead of lower() so if we wanted to add a functional index to this column, which I think we should, we'd likely do it over upper(email) so that Django's stuff works correctly.

[dcramer]

three options

• add latest_release_id to Environment?
• join Environment + Release + ReleaseCommit to select max(date_added) on commit? [ might still be too slow ]

• add latest_released_commit_id to Environment (ugh)
• Join Environment + Commit to select date_added

• Add new model which is like LatestCommit(repository_id, environment_id, commit_id)

I'm leaning towards 3 just to contain/keep it simple