Before this commit, the SQL query generated in `_get_aal_query` utilized a `LEFT JOIN` with a complex `OR` condition on the join clause: `(bl.company_id IS NULL OR bl.company_id = al.company_id)`.

Because this condition lacks a strict equality constraint, the planner cannot build a hash table for the join. Consequently, it is forced to fallback to a Nested Loop Join strategy, evaluating the condition as a filter for every row pair. This results in significant performance degradation on large datasets.

This commit optimizes the query by splitting the logic into two separate `SELECT` statements combined with a `UNION ALL`:
1. Matches where `company_id` is explicitly equal.
2. Matches where `company_id` is NULL.

By separating these conditions, the planner can now prioritize a Hash Join for the equality check and handle the NULL join separately, significantly reducing execution time.

References:
- Original PR introducing the logic: https://github.com/odoo/enterprise/pull/82955
- Plan Before (Join Filter): https://explain.dalibo.com/plan/a55476hgb73ea7g6#plan
- Plan After (Hash Cond): https://explain.dalibo.com/plan/3b9g484569a86efb#plan

opw-5460862

Forward-Port-Of: odoo/enterprise#104531
Forward-Port-Of: odoo/enterprise#104299