ENH: Adding a helper class to hold duplicated code computing p/xs offset for Pleiades sensor bundles

BUG: #942 There is no reason why the origin coordinates should be (0.5,0.5) when there is no geotransform. This induces an error when calling sensor models (which assume origin is (0,0)

BUG: Multiple errors in Pleiades sensor model (invalid cast to Int32 and inconsistency with OSSIM sensor model convention, which is that the coordinate of the UL pixel center is (0,0)

ENH: remove warning when phr mode is activated by the application and extend long description to describe app modes

BUG: Fixing #951 (applications does not work properly with external ITK) by replacing the ITK filter by an OTB one (which derives from our patched itk::UnaryFunctorImageFilter)

Change the shift apply in X and Y direction from 3/2 pixels to -3/2 pixels in
bundletoperfectsensor application for phr data. This
patch was provided by Julien Michel as he oberves that it improves the
fusion on PHR data.
I confirm that inverting this offset improve the result on PRIMARY product
samples over Melbourne.

It is still not clear for me if we've got now the "right" formula.

ENH: #942: rectify test so that leTvSOMClassifier and leTvSOMImageClassificationFilter can share the same baseline

Following recent commits on the ITK side which use clang-modernize to insert
missing override keyword, cf:

https://github.com/InsightSoftwareConsortium/ITK/commit/10757725be9cbe68ab2903c9cd4b76b738070e63

I had to compile OTB trunk using clang and  c++11 flags:

mkdir OTB-newclang
cd OTB-newclang/
CC=/usr/bin/clang CXX=/usr/bin/clang++ CXXFLAGS="-std=c++11 -stdlib=libstdc++" ccmake -DCMAKE_EXPORT_COMPILE_COMMANDS:BOOL=ON ../../../src/OTB

Internal libsiftfast does not compile as clang report error "error: argument evaluates to XXX
which cannot be narrowed to type YYY"

I added explicit cast to int in that cases.