Optimize clean_node_table's query

[AmineKhaldi]

Before: real 37.100 user 21.069570 sys 16.029508 (87118718474 cycles)
After: real 23.564 user 18.518647 sys 5.043768 (53341553604 cycles)

Before:

QUERY PLAN
|--USING INDEX sqlite_autoindex_ancestors_1 FOR IN-OPERATOR
|--LIST SUBQUERY 5
|  |--MATERIALIZE pending_nodes
|  |  |--SETUP
|  |  |  `--SCAN root
|  |  `--RECURSIVE STEP
|  |     `--COMPOUND QUERY
|  |        |--LEFT-MOST SUBQUERY
|  |        |  |--SCAN pn
|  |        |  `--SEARCH n USING INDEX sqlite_autoindex_node_1 (hash=?)
|  |        `--UNION ALL
|  |           |--SCAN pn
|  |           `--SEARCH n USING INDEX sqlite_autoindex_node_1 (hash=?)
|  `--SCAN pending_nodes
`--SCAN node

After:

QUERY PLAN
|--LIST SUBQUERY 4
|  |--MATERIALIZE pending_nodes
|  |  |--SETUP
|  |  |  `--SCAN root
|  |  `--RECURSIVE STEP
|  |     `--COMPOUND QUERY
|  |        |--LEFT-MOST SUBQUERY
|  |        |  |--SCAN pn
|  |        |  `--SEARCH n USING INDEX sqlite_autoindex_node_1 (hash=?)
|  |        `--UNION ALL
|  |           |--SCAN pn
|  |           `--SEARCH n USING INDEX sqlite_autoindex_node_1 (hash=?)
|  |--SCAN n USING COVERING INDEX sqlite_autoindex_node_1
|  |--SEARCH a USING COVERING INDEX sqlite_autoindex_ancestors_1 (hash=?) LEFT-JOIN
|  |--CREATE AUTOMATIC INDEX ON pending_nodes(hash)
|  |--BLOOM FILTER ON pn (hash=?)
|  `--SEARCH pn USING AUTOMATIC COVERING INDEX (hash=?) LEFT-JOIN
`--SEARCH node USING COVERING INDEX sqlite_autoindex_node_1 (hash=?)

[fchirica]

They look completely equivalent to me (and the green CI is a good guarantee of this).