Adding simple cubic spline algorithm for NumA++

In NumA++:
- Correct the default absolute tolerance to 1e-3 in Integrator1D, instead of 0.

In NumA++:
- Correct the warning in DExp.

In NumA++:
- Initialize Eigen vectors to zero in LSMRSolver, to avoid random results (and maybe nan).

In PARTONS_Fits:
- Propagate changes in parameters names for DVCSConstantCFFModel.

In NumA++ (trunk):
- Removed the function solveLSMR in LinAlgUtils.
- New class LSMRSolver (using Eigen vectors and matrices for the computation).
- Removed old LSMR classes.
- Various conversions of stringstream to Formatter and runtime_error to CustomException.

In NumA++_Test (trunk):
- Propagation.

In NumA++ (trunk):
- Daniele's new DExp layer.

In NumA++ (trunk):
- New LSMR ported from scipy.

In NumA++ (trunk):
- Daniele's new DExp routine.

In ElementaryUtils (trunk):
- Change GenericType and Formatter precision to 16 instead of 17, to avoid messy numbers in formatting.

In NumA++ (trunk):
- Change some std::runtime_error to CustonException. A lot more to change...
- Time monitoring in LSMR.

In PARTONS:
- Inherit IncompleteGPDModule from GPDModule to be able to use GPDService. Lot of useless code commented.
- Some changes in Radon inversion modules...

In NumA++:
- Added -pedantic flag in CMakeLists for warnings.
- Moved LinAlgUtils to parent folder.
- Added LSMR iterative algorithm for sparse least-squares problems, taken from https://github.com/tvercaut/LSQR-cpp (APACHE licence) and adapted to fit into NumA.

In NumA++_Test:
- Added -pedantic flag in CMakeLists for warnings.
- Added test for LSMR also taken from github.

In PARTONS:
- Added -pedantic flag in CMakeLists for warnings.
- Converted one observable scenario to the new cross configuration.
- Various modifications and additions to the Radon Inverse files. Now, the DoubleDistribution modules are not needed anymore. It was too complicated to deal with... But it's still possible to use them I guess.
- Renamed the Radon Inverse daughter class because the brute-force approach was dropped in favor of a fixed number of rows for the matrix.

In PARTONS_exe:
- Added -pedantic flag in CMakeLists for warnings.
- Updated partons.properties to the new scheme.
- Added a test for the Radon inversion.

In PARTONS_release:
- Added -pedantic flag in CMakeLists for warnings.
- Added a scenario for GPD.
- Close PARTONS properly when exception.

In NumA++ (trunk) :
- Replace std::runtime_error by CustomException.

In PARTONS:
- Updates to Radon module

In NumA++:
- Remove blank files (created to avoid svn problem).

In ElementaryUtils:
- Use command "executable" in the Ant scripts instead of the deprecated command "command".

In NumA++:
- Overload functors to allow call without parameters.
- Restructuring of folders for Functors.
- Rename LinearSystem to LinAlgUtils and add rank routine.
- New EigenUtils to encapsulate conversions to or from Eigen. In order for PARTONS to be independent of Eigen.
- New methods in MatrixD (appendLine, etc).
- New constructors for VectorD.
- New static method in Interval to compute nodes.

In NumA++_Test:
- New tests for Eigen.

In PARTONS:
- New modules IncompleteGPDModule and RadonInverseModule.
- First method for Radon inversion: Random brute force with constant piecewise functions (RandBFConstPW). Still no inversion, only matrix construction.
- Addition of DDGauge type and modification of DDType.
- New OverlapMMR15 model.

In NumA++ (trunk):
- in linear_algebra/matrix_complex and linear_algebra/vector_complex
	- dynamic vectors/matrices replaced by std::vectors
	- documentation added
	- memory leakage fixed
	- few cosmetic changes

In NumA++ (trunk) :
- Create FunctorUtils class : add missing method to create FunctorMD.

In NumA++ (trunk) :
- Add multidimentional functor.

In NumA++ (trunk) :
- Update eclipse project configuration file.

- Use command "executable" in the Ant scripts instead of the deprecated command "command".

In NumA++, following modification in PARTONS_Fits:

- Update FindROOT.cmake module for a finer resolution of ROOT components in system. Needed because previous version was not able to find libMinuit2.so

- Update build.xml: remove ROOT home path argument.

New classes that generates quadratures for numerical integration.

Chebyshev A: Chebyshev quadrature for computing the integral \int_{-1}^1 dx f(x)sqrt(1-x^2);
Chebyshev B: Chebyshev quadrature for computing the integral \int_{-1}^1 dx f(x)/sqrt(1-x^2);
GLSeSt: Different (more precise?) method for computing GL weights. 

In NumA++ create trunk.

In ElementaryUtils:
- Added string split function by string delimiter in StringUtils.

In PARTONS:
- Changed ObservableKinematic to use GenericType instead of fromStringToDouble.
- Some fix in DVCSConstantModel.

In PARTONs_Tests:
- Force C++98 in CMakeLists.

In NumA++ :
- Implement Registry/Factory (as a draft) for Integrator.

In NumA++ :
- Implement Registry/Factory (as a draft) for Integrator.

NumA++ : add .project file needed by eclipse autoconfiguration tool.

- Fix CMakeLists.txt to force C++98
- Fix some absolute URL

- NumA++ : create tag for r686

Initial

Initial

Revert branches

New Nabil branch

- NumA++ : create release branch

- NumA++ : update SVN folders tree for trunk and branches

- NumA++ : remove default SVN root trunk.

- NumA++ : update SVN folders tree for trunk and branches

- NumA++ : update SVN folders tree for trunk and branches

- Add comments