## [FIX] account: sequence mixin cache on cr.cache instead of precommit
The aim of this commit is to prevent the creation of Savepoint in chain.
This is achieved by preventing the `sequence.mixin` cache to be cleared
as soon as a flush happens and to prevent the `bank.statement.line`
creation to clear the cache.

### Context:
During an import of document, like a big CSV of bank statement for
example, if any issue arises, the ORM rollback to the savepoint of
import and retry to import one record at a time and rollback again
later on to collect the errors and show the faulting line to the
customer.

### Cause:
1) As we flush for every single record, the `precommit.data.cache` gets
   cleared during the flush and the `sequence.mixin` cache gets wiped
   out.
2) The creation of bank statement set the name of its move_id to False
   which results in the `sequence.mixin` code to clear its own cache by
   itself.
   [name set to False](https://github.com/odoo/odoo/blob/1361c0bc98c91f1e601b0fc3beaa948df3a3bfcd/addons/account/models/account_bank_statement_line.py#L410-L412)

Both (1) and (2) results in the sequence.mixin code to recreate the
cache and, for that, to create a new Savepoint.
This is done at each iteration of the import loop which could results
in this case in 300+ savepoints existing at the same time eating all
Postgres shared buffers memory putting the whole production database
on its knees.

task-id: None (investigated for odoo.com)

Forward-Port-Of: odoo/odoo#250827