Possibly useless Spanner work

[data-sync-user]

I was looking at Spanner's query insights and I see the query using the most CPU time in the last [whatever period of time you choose] is:

SELECT COUNT(*)
            FROM bsos
           WHERE fxa_uid = @fxa_uid
             AND fxa_kid = @fxa_kid
             AND collection_id = @collection_id
           GROUP BY fxa_uid

It uses about 18% of total CPU, according to query insights (attached image).
I looked for that query in the code and only found it here: 

syncstorage-rs/syncstorage/src/db/spanner/models.rs

Lines 870 to 875 in a52900f

	"SELECT COUNT(*)
	FROM bsos
	WHERE fxa_uid = @fxa_uid
	AND fxa_kid = @fxa_kid
	AND collection_id = @collection_id
	GROUP BY fxa_uid"

It looks to me like it runs if quota is not enabled (and we just talked about how that was a feature we never turned on).

And ALSO, it looks like the results of that query are NEVER USED if quota is not turned on (see: 

syncstorage-rs/syncstorage/src/db/spanner/models.rs

Lines 920 to 926 in a52900f

	} else {
	"UPDATE user_collections
	SET modified = @modified
	WHERE fxa_uid = @fxa_uid
	AND fxa_kid = @fxa_kid
	AND collection_id = @collection_id"
	}

)So it seems to me like we could save a lot of (Spanner CPU) time by removing that query altogether.

After a slack convo with @jr Conlin he pointed out that if there are no results, then other things happen, so this query is important, but only in a “are there zero rows or more” sense. So I propose changing the query to something along the lines of:

SELECT 1
FROM bsos
WHERE fxa_uid = @fxa_uid
AND fxa_kid = @fxa_kid
AND collection_id = @collection_id
LIMIT 1

Returns 1 or [no results]. And shouldn’t have to scan a bunch of rows

┆Issue is synchronized with this Jira Task