Replenishment report no longer scales cubically

Commit message

[PERF] stock: speed up orderpoint action

Before this commit, calls to `is_parent_path_in` when forecasting
quantities scaled cubicly with locations x products x quants/moves. A
lot of work is also done on product locations with an existing
orderpoint, all to be discarded at the end. The line
`orderpoint.qty_forecast += product_qty` has no tangible effect since
the field is non-stored with no inverse.

This commit lifts the quantity computation to be directly from the
quants/moves data, and stores the result in a dictionary to be accessed
later. This also elminates the need for uom float comparisons for
products without quants/moves.

This commit also removes all unnecessary intermediate steps for product
locations which already have an orderpoint.

This commit also applies other trivial optimzations:
- Remove intermediate recordset when calling
  `_unlink_processed_orderpoints`
- Order by id on searches where the default order is not needed.

Benchmarks
---
+-----------------------------+------------+---------------+
| Scenario                    | Total Time | Total Queries |
+-----------------------------+------------+---------------+
| 18      Replenish Locations | 17s        | 1,175         |
| 31,000  Replenish Products  |            |               |
| 141,000 Quants              | After:     | After:        |
| 14,000  Moves               | 3.64s      | 634           |
| 472,000 Orderpoints         |            |               |
+-----------------------------+------------+---------------+
| 35      Replenish Locations | 389s       | >50,000       |
| 89,000  Replenish Products  |            |               |
| 298,000 Quants              | After:     | After:        |
| 539,000 Moves               | 19s        | 1,068         |
| 634,000 Orderpoints         |            |               |
+-----------------------------+------------+---------------+

opw-6357118