Rewrite cluster status query to only get unique hosts

[nothingface0]

Previously it was grouping by the is_up column too, returning duplicate host entries.

The query was rewritten to work as intended:

SELECT h.name, t.is_up, t.message, t.created_at
FROM hoststatuses t
JOIN (
    SELECT host_id, MAX(created_at) AS max_created_at
    FROM hoststatuses
    GROUP BY host_id
) sub ON t.host_id = sub.host_id AND t.created_at = sub.max_created_at
JOIN hostnames h ON t.host_id = h.id;

A multicolumn index on host_id and created_at was created for the hoststatuses table to increase performance.

Queries now take ~150 ms (compare to the ~220 ms it took before).