Optimize language existant query

[bjester]

Summary

The following query-- which originates from the publish modal providing a list of content languages, or from testing the existence of a language in a channel-- was taking too long:

SELECT DISTINCT "contentcuration_contentnode"."language_id",
                "contentcuration_contentnode"."tree_id",
                "contentcuration_contentnode"."lft"
FROM "contentcuration_contentnode"
WHERE (
    "contentcuration_contentnode"."language_id" IS NOT NULL
    AND CAST("contentcuration_contentnode"."node_id" AS uuid) IN (
        SELECT U0."id"
        FROM "kolibri_public_contentnode" U0
        WHERE U0."channel_id" = '<CHANNEL_ID>')
            AND NOT ("contentcuration_contentnode"."id" IN ('<MAIN_TREE_ID>'))
)
ORDER BY
    "contentcuration_contentnode"."tree_id" ASC,
    "contentcuration_contentnode"."lft" ASC
;

The following issues were addressed with the query:

• Removed querying against the public nodes since those are only populated after publishing and for public library channels
• Removed the default ordering added by MPTT
• Updated the query to directly query against the content nodes using the tree_id then filter from there

The updates should now produce a SQL query similar to:

WITH t AS (
  SELECT id, language_id
  FROM contentcuration_contentnode
  WHERE tree_id = '<TREE_ID>'
)
SELECT DISTINCT t.language_id
FROM t
WHERE t.language_id IS NOT NULL
  AND t.id != '<MAIN_TREE_ID>'

References

#4666
https://learningequality.slack.com/archives/C0LK8QS9J/p1733174857751199

Reviewer guidance

• Tests were implemented for the original feature, which were ran before and after, and passed each time

[akolson]

Changes make sense to me and query is now more direct. Thanks @bjester for the optimization.