This test (just like all legacy filesystem tests) does not use business
template dependencies but just install business templates in the order from
getBusinessTemplateList, but business templates are not listed in correct
order.

It use to be OK (which warnings), but erp5_test_result can no longer be
installed without erp5_project, so review installation order to install
erp5_project earlier.

See merge request nexedi/erp5!1222

Tabular Data Tool is a Tool to handle tabular data (2-dimentional table data)
in erp5.
The tabular data type is currently a plain pandas DataFrame. It is intended to
use with ERP5 Reports to create spread sheets (ODS/Excel).

With Tabular Data Tool, we can do those things on reporting:

- transform dict-list, portal-catalog-result and document-list into a tabular data
- joining/grouping/sorting the tabular data
- using vectorizing function to optimise time consuming reports

Catalog seems to correctly handle the OR parameters.

Do not lose the possible external `query` parameter.

Test jumping to the same context.

Test also direct form rendering configured as jump action.

This allows getting stable URL which can be used in ERP5JS to highlight the current jump

This test uses pinDateTime to create an history of test results where tests
have been executed in order test suite 1, 2, 3.
This is done using this pattern:

      now = DateTime()
      ... # do several things that may actually take more than one second
      self.processTest("test suite 1", "r0=a", node_title="COMP0-Node1")
      self.pinDateTime(now + 1.0/86400)
      self.processTest("test suite 2", "r0=a", node_title="COMP1-Node1")
      self.pinDateTime(now + 2.0/86400)
      self.processTest("test suite 3", "r0=a", node_title="COMP1-Node1")

My guess is that we take "now" too early, and when we add 1 second to this
now variable, this may actually be going back in time sometimes, effectively
creating test suite 2 before test suite 1.

Assign now to current date after doing all the affectTestSuite calls should
solve this.

Portal Type as Classes: Interaction Workflows methods could be improperly generated because of 58d4ab8e.

1. Add Interaction on _setSourceReference() to `Document Component`.
2. generatePortalTypeClass() is called recursively and an inner `Document
   Component` Portal Type class is created without any accessor holder.
=> As __bases__ is (klass, GetAcquireLocalRolesMixIn) since 58d4ab8e but loadClass()
   was not updated to reflect this, Portal Type Workflow Methods were generated on
   the inner Portal Type Class where _setSourceReference() is not available and
   registerWorkflowMethod() creates a WorkflowMethod calling Base._doNothing.

The performance of a module is much less impacted by the number
of objects than before, at least for kind of total we have in
tests (about 1000 objects)

Round the amounts line by line when validating transactions balances and
showing the totals on an accounting transaction.

Even though float field configurations do not allow users to input transactions
where amounts uses more digits that what makes sense for the selected currency,
we had cases where some transactions generated by custom scripts were "too
precise" and passed the validation because unlike in all reports where we
display sum of rounded values, the check was only rounding the sum, which cause
some transactions with 1.00 on one side and 0.333333 0.333333 and 0.333333 on
the other side to be valid.

The scripts used to display sums on an accounting transaction had the same
problem, they were also showing the rounded sum and not the sum of rounded
values.

Always use textwrap.dedent implicitly so that we don't have to do it in
every test and disable the way of making multi lines scripts by passing
multiple arguments. Now there's only one way to do and every tests use the
same way

getRepositoryList contain all business template repositories, ie:

      ['/srv/slapgrid/slappart4/srv/slapos/soft/c63ba7265399450b28f9ea6d5667a5e7/parts/erp5-doc/bt5',
       '/srv/slapgrid/slappart4/srv/slapos/soft/c63ba7265399450b28f9ea6d5667a5e7/parts/erp5/bt5',
       '/srv/slapgrid/slappart4/srv/slapos/soft/c63ba7265399450b28f9ea6d5667a5e7/parts/erp5/product/ERP5/bootstrap',
       '/srv/slapgrid/slappart4/srv/slapos/soft/c63ba7265399450b28f9ea6d5667a5e7/parts/erp5-bin/bt5']

it seems in random order.

Test was written to use the one from parts/erp5/bt5, but since it just
takes "anything which does not contain bootstrap" it was sometimes getting
erp5-bin or erp5-doc and could not find erp5_configurator_standard business
template.

Change the selection of the repository to take into account that we may also
have erp5-doc and erp5-bin and so that it fails if more than one.

all member of random module are available (because in ZopeGuards, there's a
random.__allow_access_to_unprotected_subobjects__ = 1), SystemRandom class
also is, but Random class was not, even though there should be nothing unsafe
in this.

This is needed if we want to use a random number generator with a fixed seed
to get reproductible results. We could use random.seed module level function
but since this is also used by several things in ERP5 this does not produce
reproductible sequences.

See merge request nexedi/erp5!1265

ZODB Components: TestZodbDocumentComponentReload (e5587493): No need to run other common Component tests.

Before we had the sub-documents exported as properties of container document
for some reason. Now they are exported as separate documents, as expected.