Implementation dependant parameters, such as message type or quality,
must be defined by the gateway class, not in the common
ShortMessage_send script.

This should make it easier to switch implementations.

Also:
 * send method is expected to be called once per recipient. I guess we
   prefer to isolate messages than to batch send messages efficiently.
 * send method expect relative url of sender and reciever documents, so
   that it can decide which properties to use.
 * drop `sender_title` parameter, now the gateway implementation decide
   what to use.
 * drop `test` parameter. For simulation mode, set simulation mode on
   the gateway instance
 * sms.send() no longer accept direct from_url / to_url / body message.
   It only supports sending what's defined on document properties.
 * SMSTool_afterSend will be called with message_id parameter, not
   message_id_list
 * Split interfaces in two SMS Sending / SMS Receiving.

- fix message_type to support multitext messages
- use MULTITEXT message type by default

Only mobyt support this parameter, so this break the compatibility with
essendex

Before this change, expression_instance would not be set when expression is
modified, leading to either non-applied filter expression (when no
expression_instance existed) or use of an outdated expression.
Use a volatile attribute to store the expression: it does not change the
amount of work needed to load the Expression instance (pickle is the raw
python expression as text) and it remove data duplication in exported
documents.

This reverts commit fe84b799.
I did not intend to push that hash (copy/paste mistake).

monospace, sans-serif, serif, OpenSans

[hal_json] Improve backward compatibility of value resolution by calling Scripts on search result underlying object

This adds the maximize button.

This reverts commit c49cf8d8.
Callee is directly accessed from other places, so changing this API is
wrong.
This it not immediately visible in tests, but this makes
erp5_officejs_support_request_ui_test:testFunctionalSupportRequest
testCheckComment test randomly fail (apparently because the javascript
error message may appear after or before the test final step is
satisfied - which means this tests is bugged, but that's a separate
issue).

These variables are only used when they evaluate as true, so it is
pointless work to replace them with empty lists.

Catalog used in this test is now close/above 1k rows, so go one API level
down to get control over that pesky "limit 1000".

These tests do not test what they advertise anymore. Advertised features
are in widespread use in other tests, so there should be no need to test
specifically for these.
Also, avoids polluting other tests by clearing catalog for futile reasons.

ZSQLCatalog/ERP5Catalog is only supposed to work on catalog table, by
definition.

Also, reorder imports to follow python style: standard modules before
others.