initialize the distributions on the GPU

[atmyers]

Parallelize particle initialization using ParallelForRNG for all backends.

Follow-up to #73
Close #495

• ensure GPU results are correct

[ax3l]

ImpactX support MR now. I wonder if lev!=0 is generally needed or if we want to remove the parameter altogether.

I assume that on Redistribute we automatically move the particles in the corresponding level at the end of AddNParticles?

[ax3l]

Ok, this passed the thermal tests on GPU on Perlmutter 🎉

[ax3l]

@cemitch99 a couple of tests fail with this PR on GPU due to tolerances in analysis scripts. The main reason is that we change the way we draw random numbers and have different seeds in parallel execution now.

Unless there is a GPU bug, might need to relax tolerances a bit, I think.

[ax3l]

The following tests FAILED:
	  2 - FODO.analysis (Failed)
	  5 - FODO.MPI.analysis (Failed)
	  8 - FODO.py.analysis (Failed)
	 11 - FODO.MADX.py.analysis (Failed)
	 14 - FODO.py.MPI.analysis (Failed)
	 17 - chicane.analysis (Failed)
	 20 - chicane.py.analysis (Failed)
	 23 - chicane.MADX.py.analysis (Failed)
	 34 - gaussian.analysis (Failed)
	 38 - FODO_RF.analysis (Failed)
	 40 - FODO_RF.py.analysis (Failed)
	 42 - kurth4d.analysis (Failed)
	 44 - semigaussian.analysis (Failed)
	 46 - multipole.analysis (Failed)
	 48 - expanding_beam.analysis (Failed)
	 50 - expanding_beam.py.analysis (Failed)
	 52 - multipole.py.analysis (Failed)
	 58 - iotalattice.MPI.analysis (Failed)
	 60 - iotalattice.py.MPI.analysis (Failed)
	 62 - kurth_periodic.analysis (Failed)
	 64 - kurth_periodic.py.analysis (Failed)
	 74 - solenoid.analysis (Failed)
	 76 - solenoid.py.analysis (Failed)
	 78 - solenoid.MADX.py.analysis (Failed)
	 88 - quadrupole_softedge.analysis (Failed)
	 90 - quadrupole_softedge.py.analysis (Failed)
	 92 - fodo_chromatic.analysis (Failed)
	 94 - fodo_chromatic.py.analysis (Failed)
	102 - cyclotron.py.analysis (Failed)
	110 - compression.py.analysis (Failed)
	136 - IOTA_lattice.analysis (Failed)
	138 - IOTA_lattice.py.analysis (Failed)
	143 - aperture.run (Failed)
	144 - aperture.analysis (Failed)
	145 - aperture.py.run (Failed)
	146 - aperture.py.analysis (Failed)
	148 - apochromat.analysis (Failed)
	152 - alignment.analysis (Failed)

[ax3l]

For the FODO cell example, the initial beam sigma_x,y and emittance_x,y is 2 orders of magnitude smaller then the reference value.
That makes no real sense to me, since we set finite numbers. Could be a GPU bug.

Update: misread, indeed just sampling noise. Will slightly push tolerances.