Fix bug when buffer_size=1 in TuneMonitor by switching to init buffer_count to 0. Now total_size should be 1 less.
Improve plot_tunedata

Fix profile monitor
Add detune to cavity monitor

Add a dict_dtype optional parameter to Monitor.monitor_init to allow to choose the dtype to use to save the data.
Improve CavityMonitor to allow for saving beam and cavity phasor for all bunches
Add a plot_cavitydata function
Fix a bug on CavityResonator.track -> nturn += 1 at correct time
Change CavityResonator.track to allow for monitoring the cavity/beam phasor at each bunch

Change back invariant name to cs_invariant
Simplify SkewQuadrupole track method

Add tracking to CavityResonator based on the mbtrack implementation but using a analytic formula for speed-up
Add method to Beam to share longitudinal distributions
Add methods to Mpi to look for next and previous bunches
Add start of a CavityMonitor, to improve to deal with multi-cores

- cs_invariant method in Bunch is renamed to 'action'.
- Action is now saved together with BunchMonitor and can be plotted in plot_bunchdata

Bugfix due to switch to numpy arrays.
Allow to save less mp than the full number in the bunch.

Bugfix2 : PhaseSpaceMonitor

Change track method to allow for Beam objects
If pynaff can not find the tune, the tune value is nan.
Nanmean and nanstd are used to avoid to bias the tune value.
! Untested if some particles are lost !

WakePotentialMonitor now use interpolation on a fixed grid to be able to save the wake potential data (which size change from turn to turn).

Remove tools.py to avoid circular import problem with wakefield.py -> merged into wakefield.py
Remove machines folder to save machine data in a seperate git instance
Rework import so the code can be used as independant module mbtrack2

Fig bug for ProfileMonitor if dimension is not in list

Add ProfileMonitor to save bunch profile
Add a track_bunch_data to Monitor class to group the code of track methods from BunchMonitor and PhaseSpaceMonitor
Implement BeamMonitor for the case without mpi
Remove hard coded h=416 ! in BeamMonitor

The HDF5 is now open only once on all the cores to be able to write in parallel in the same file. The file must be closed at the end of the tracking.
BeamMonitor is used to save global statistics of all bunches
Add docstrings
TODO : Adding a flush() call at some point might be needed !

Add a monitor module to save tracking data based on h5py
Bunch Monitor is used to get statistics from a specific bunch
PhaseSpaceMonitor is used to save the phase space of a specific bunch
For now the .hdf5 file is opened and closed at each write operation, this seems to cause some problems for mpi.