[SYNCOPE-1855] Rewriting JPAAnySearchDAO to reduce subqueries

[ilgrosso]

The current JPAAnySearchDAO implementation is based on incrementally constructing the final query string.

The new implementation is instead meant to build a tree representation and finally generating the query string from it.

The new implementation was initially provided as AnySearchNodeDAO, not updating JPAAnySearchDAO yet to avoid build errors under core/persistence-jpa-json.

TODO:

• check and fix integration tests running on standard persistence (H2, PostgreSQL, MySQL, MariaDB, Oracle)
• adjust and check the JPA JSON implementations (PostgreSQL, MySQL, MariaDB, Oracle)

[ilgrosso]

ojson current issue; the generated query is

SELECT DISTINCT sv.any_id,sv.username
FROM user_search sv
WHERE (((plainSchema='fullname' AND ustringValue LIKE '%o%' ESCAPE '\' ) 
AND ((sv.any_id IN (
SELECT DISTINCT any_id FROM user_search_umembership WHERE group_id='37d15e4c-cdc1-460b-a591-8505c8133806') 
OR sv.any_id IN (SELECT DISTINCT any_id FROM UDynGroupMembers WHERE group_id='37d15e4c-cdc1-460b-a591-8505c8133806')))) 
AND (plainSchema='loginDate' AND JSON_VALUE(dateValue, '$[*]')='2009-05-26T00:00:00+02:00') 
) 
ORDER BY sv.username ASC

but returns no results, while the amended

SELECT DISTINCT sv.any_id,sv.username
FROM user_search sv
WHERE ((plainSchema='fullname' AND ustringValue LIKE '%o%' ESCAPE '\')  
OR (plainSchema='loginDate' AND JSON_VALUE(dateValue, '$[*]')='2009-05-26T00:00:00+02:00'))
AND ((sv.any_id IN (
SELECT DISTINCT any_id FROM user_search_umembership WHERE group_id='37d15e4c-cdc1-460b-a591-8505c8133806') 
OR sv.any_id IN (SELECT DISTINCT any_id FROM UDynGroupMembers WHERE group_id='37d15e4c-cdc1-460b-a591-8505c8133806'))) 
GROUP BY sv.any_id,sv.username
HAVING COUNT(*) = 2
ORDER BY sv.username ASC

works as expected.

The problem is recognizing sibling AND conditions for plainSchema and refactoring as OR conditions with GROUP BY / HAVING.