Fix a large bunch of warnings, and there are still mistakes that should be fixed before the SC release

At the user levl, we wait for a sequence. We need to add a MORSE_Barrier function that waits for all sequences currently running

instead of introducing RUNTIME_distributed_barrier

add cudablas library to make calls to cuda kernels (magma here, cublas will follow), no more calls to magma in runtime/starpu codelets

change the way we include our own header files --> relative to the root - when plasma is in the same env, chameleon can take some headers not belonging to it (ex: #include descriptor.h, this file states in plasma install dir also) which make compilation errors

change installation directories for headers, executables and docs. Make it relative to chameleon to avoid a bloody mess in system dirs

MORSE_Desc_Create_User with NULL pointer cannot be used to allocate the data (mat) - when NULL is given to MORSE_Desc_Create_User means that the application manage the allocation - to let MORSE allocate its data -> use MORSE_Desc_Create with NULL pointer

change copyright - correct whitespace - place cmake module depending on chameleon in cmake_modules and no more in cmake_modules/morse