Make the M2L non-idle during the communication, but it relies on taks (with chunck size) otherwise nested parallelism will be needed...

Now the interval of workers is max+1;max if they have no more work on a level (and indent the whole file)

remove warnings  in FFmaGEnericLoader (sign vs unsigned )

Works fine until 4 process MPI, with 5 processes, the total amount of energy decays. Any number of thread should works.

Hybrid Algorithm modified, and Parallel Details paper updated, with some tips for using EZTrace on ScalFMM

Changes to FFmmAlgorithmThreadProc and Spherical Cell in order to compile with Intel without having any problem with the GetSize function

Chebyshev and Uniform kernels can be used now with distributed algorithm. 2 test files have been added to use these kernels with FmmAlgorithmThreadProc

MpiBuffer works fine, a test file using MPI and Rotation kernel has been created in Tests/Utils, FBufferWriter/Reader is no longer used in testFmmAlgorithmProc

FMpiBuffer reader and writer are ok. There are only used by FFmmAlgorithmThreadProc :: M2M for the test. Changes will be spread out everywhere.