[REF] stock: perf orderpoint as batch

Currently orderpoint run one request at a time. However it's not
efficient since compute are some private methods could be executed
one time in batch instead of record by record.

The strategy in order to improve it is:
- Still cut the orderpoints in batch of 1000 (usefull when using a new
cursor with intermediate commit)
- For each batch of 1000, create a list of procurement values and run
them in batch (commit a2a6b099a4c1c99f4aa9e741442944e588a8f048 for more info)
- Each time a procurement exception is raised, catch the failing
orderpoint and rollback to the state before run. Remove the failing
order point from the current set and run it again.
- run method try to catch a maximum  errors by continuing to process even it
finds error in order to stack them. It would avoid to run the same set
again and again each time an error is raised.

test_orderpoint_1 took 975s before this patch and 100s after.

Task: 1849912

0 files changed, 0 insertions, 0 deletions