Change tpgqrt interface because ungqr already includes laset, so it can not be used as a two steps algorithms

    - use (starpu_cpu_func_t) 1 trick, same as cuda_func
    - cpu funtions are not defined anymore avoiding the dependency to coreblas
    - add #if !defined(CHAMELEON_SIMULATION) where it is needed
    - remove dependency to the coreblas library (become useless)
    - remove useless simucblas, simulapacke libraries
    - remove CHAMELEON_SIMULATION_MAGMA cmake variable and definition
      - keep using CHAMELEON_USE_CUDA and CHAMELEON_USE_MAGMA to consider CUDA kernels
      - this avoid to introduce useless new variables
    - work on messages

fix use of cuda+magma in simulation mode: size of workspace must be consistent between real and simu modes

The new codelet is added for all 3 runtimes but only tested with starpu ;-)
An example test7 is added : it is a copy of test6 that uses the new function to build the matrices