- reduce probability and cost of ConflictError by posting the
  activity result by activity
- do not post None (to force it, it's still possible to forge an
  ActiveResult object containing the None value)
- allow grouped messages to post results (previous implementation was not
  working, and this required to change the API of grouping methods)

More changes should be done so that Activity Tool provides an API to activate
an object using only its path.

It defaults to .01 and must be in the interval ]0, 1]
If a message expands to N objects, then it costs N / group_method_cost
and activity tool stop reserving messages the sum of all costs exceeds 1.

By default, this preserves original behaviour, i.e. MAX_GROUPED_OBJECTS = 100

Value 1 is not rejected to allow generic coding
(ex: activity_kw['group_method_cost'] = 1. / packet_size)

Tester accessors was not generated for acquired properties,
only content accessor was generated.

Extend test to prove the fix.

This Configurator Item was a Legacy which has no importance anymore, as
Business Template is not exported and installed on third party Instances
during the configuration.

The change from ocropus to tesseract as the conversor application
responsable  to convert image to text, made this tinny difference
between the outputs.

The difference was manually tested by Rafael to verify that it was caused
by the change of conversor application.

The Telephone is always presented using "+(0)" format, this is a feature
and expected behaviour.

is set or will be set later (through a Credential Recovery).

- Use _[ds]etOb instead of _[ds]etObject and unindex old object only if
  required. This improves performance and workarounds a bug documented in
  test_02_unindexObjectDependency.
- Automatically fix 'portal_type' attribute if required.

Return result of redirect call as it's used
in testERP5Credential.

- Avoid overhead of getPropertyList & getXxx accessors
- Acquisition is useless for 'specialise' category on Simulation Movements.
  Maybe the same optimization can be done for all categories ?

Just a small performance improvement, the behavior still the same.
Because when the client send a big file, it should not take too long to check the sha512 sum.

Shacache instance is now able to provide big file without getting killed
by the OS, due a memory leak.

Desactivating the previous object is an approach to send the data without
overloading the RAM Memory.