This export action produce a spreadsheet containing information needed to physically prepare sales packing lists, prepare a mailing to prepare address stickers or whatever specific processing "outside of ERP5" can be needed when sending sales packing lists.

It is an action on sales packing lists module:

![Screenshot_2017-07-10_at_14.30.10](/uploads/554a0d7b656aaa88d250136dd39309f3/Screenshot_2017-07-10_at_14.30.10.png)

Offering two level of details:
 * One line per sales packing list line
 * One line per delivery, with the resource / quantities "packed" in a column

Both versions will display a list of sales packing lists, but only the one where resource is a product. The idea is to only care only about physically delivered goods.

When requesting boths (which is the default), user will get a spreadsheet with two tabs, like this example: [Shipment_Report.xlsx](/uploads/c39906f36cbddd5c20cdd4ce39a90381/Shipment_Report.xlsx) (this is using data from `testTradeReports.TestTradeReports.testShipmentReport` added in this MR)

The intended use case of this action is to select packing lists to prepare (usually with worklist), export this spreadsheet and change state of the selection.

/reviewed-on nexedi/erp5!320

/reviewed-on nexedi/erp5!345

/reviewed-on nexedi/erp5!343

/reviewed-on nexedi/erp5!344

/reviewed-on !340

/reviewed-on nexedi/erp5!341

- fix `NoneType` as no attribute `getTitle` (service_obj.getTitle()) ;
- use existing variable as most as possible (portal, result) ;
- avoid call `getTitle` on the same object in a loop (project.getTitle()) ;
- avoid compare projects by title as many projects may have the same ;
- avoid using `set` as service id cannot be present twice (set(service_pairs)) ;
- optimise catalog request to get project (limit) ;
- optimise catalog request to get sale supply (destination_project_uid) ;
- rename plurals in variable (service_pairs) ;
- rename abbreviations in variable (*_obj) ;
- remove trailing spaces.

[renderjs_ui] testSortItemAutoBuild counts with strict sort editor and does not fail on fast computer

As it seems addToDate do not just add 1 unit of time, but is sensitive
to the number of days in a month too

even if there is only one workflow action

/reviewed-on nexedi/erp5!335

According to Payzen API, boolean should be translated into 0 or 1 in order to be part of the signature.
The only value which contains a boolean is "litige" which can cause signature miss-match without
this change.

  Reimplement for follow up the migration of PropertySheets from Products to ERP5

  Normally it should work as annonymous.

the selection name is duplicated with Computer_view/listbox

It can be used with an external method by providing the URL to proxy as a query parameter:
ERP5Site_getHTTPResource?url=URL

The query is synchronous (no portal_activities), so, the Zope thread will be blocked until the external request is other.
In order to prevent consuming all threads, there is a really aggressive timeout of 1 second only.
This is enough for a RSS reader.
If the timeout is to short, do not increase it, but change the proxy design to use portal_activities instead.

For now, only authenticated user can use it (to prevent becoming an open proxy).
Only GET query are allowed (until a use case required any other HTTP method).

Instead of using obsolete processing_node parameter, use invalid_onlym
with True, in order to check if there are activities with error.

As INVOKE_ERROR_STATE is needed only in SQLBase, move it there and so drop it
from ActiveObject and ActiveProcess.

Thanks to using invalid_only the caller can check if there is any
invalid activity on the context.

This might be used by hasErrorActivity in ActiveObject and
ActiveProcess.

hasErrorActivity is tested on ActiveObject and ActiveProcess classes.

DISTRIBUTABLE_STATE was removed bad50097

Writing SQL belongs to catalog.

Improvement: As proposed by Jean-Paul when this extension was added (2010),
allows mixing left- and inner-join lookups.
Improvement: Now it is not required to specify tested_base_category_list
to enable inner joins: they will automatically be used whenever a base
category set on context matches the preference. If all categories set on
context match the preference, then no left join will be used at all.

This only makes a difference when "test=False" is given, which is a hack
anyway.

The main purpose is to stop calling buildSQLSelector so it can be removed.
Remove Base_zSearchRelatedObjectsByCategoryList.
Also, discard a superfluous comment in testArchive which references the
discarded ZSQLMethod, for consistency.
Fixes an existing failure in testArchive.

Less ambiguous than previous syntax while keeping names short.
Allows defining more flags without limiting base_category namespace: the
only forbidden base_category id is "related", which was already reserved
(ie, not working) in previous syntax.