* CMakeLists.txt: increment version number and update dependency
	  requirements.

	* src/SimplePluginManager.cpp: aBuffer was allocated with size 32768 and
	  initialized to 0 over "size" bytes. Sometimes, "size" haapens to be
	  more than 32768.

Because there are now several instances of Preview control
with different settings, this commit draw a clean cut
between Kajita's scheme and the others regarding this object.

Recursive calls to the weights seems to trigger different
resolutions leading finally to the apparitions of Nan.
This commit is just for book-keeping.

Add a sensitivity term to the generation of the polynomial.
Due to the impossibility to represent some numbers, the strict
test on time value fails and some value for the feet may not be
generated.
This happen quite frequently for long sequences of steps.

Previously the computation of the weights had to be triggered manually.
This is not anymore the case with this commit.

Inherited method is not called. This can be fix by a better inheritance/protected mecanism
(to be implemented soon).

Following complaints from users on the downloading time,
those files are removed. Thus it will still be long to
clone the repository but no extension of the time is
to be expected if the documentation is changed.
Moreover the next documentation will be limited
to the public interface, generated by doxygen.
For people interested by the algorithms, please
read the papers given in reference in each source file.

	When m_SamplingPeriod=0, the program may exit whitout segv when trying
	to resize a vector with a dummy size.

Call EvaluateStartingState in CommonInitialization.
Closer to SoT setup.

Deals with wrong parameters initialization when
calling PreviewControl::ComputeOptimalWeights.

Plus code cleaning up.

The code is clean of reference to HRP-2 or
related to specific code of GRX.

Morisawa:Validated through experiments and the Stack of Tasks
Kajita:Validation through OpenHRP-2 dynamic (seems unstable)
not yet in the robot.
Wieber2006: Untested.
Dimitar2009: Untested.

DoubleStagePreviewControlStrategy was unable to read properly
the previewcontroltime and samplingperiod calls.
Fix a problem in the delayed footsteps introduced in the
FIFO.

The initialization should compute the position of the CoM
in the reference frame for which the center is the projection
of the Waist on the floor.

The FIFO of the feet is now properly
synchronized with the FIFO ZMP reference.
Code cleaning removing old debbuging part.