separate buffers for different particles (SOAOSOA?) #309

This is about the second issue discussed in #175 (comment)

Eventually, it may better to change the buffer indexing so as to have completely separate buffers for different particles. This would simplify the ixxx/oxxx api (no need to pass ipar) and make the code more understandbale. However it is clearly something to be synced with the fortran bridge work. Not urgent.

One possibility could be to move from the current AOSOA to an SOAOSOA (or back to an SOA):

• now we have momenta as AOSOA[npagM][npar][np4][neppM}, with npar particles, np4=4 components of 4-momenta, and nevt=npagM*neppM events split into npagM pages of neppM events per page
• eventually we could have SOAOSOA[npar][npagM][np4][neppM], so that it is easy to split these into several separate AOSOA[npagM][np4][neppM]
• or we could use SOA[npar][np4][nevt=npagM*neppM], although some tests I had done long ago seems to say there was a penalty f rthis (tests to be repeated)

Anyway, it could be useful to investigate this