Multiple hdf5 file can now be made without having to use MPI. This is achieved by clearing the file name storage every time the monitor is closed.

New LongRangeResistiveWall class to deal with multi-bunch and multi-turn wakes from resistive wall.
Add Mpi.share_means method to share bunch means using MPI.

Fix error on plot_beamdata.
Add option to specify ylim to streak_bunchspectrum and to streak_beamspectrum.

Fix TD calculation for transverse dipolar wakes due to sign change in the 31/08/2021 commit.

Add a is_empty property to the Bunch class.
Add a check_empty optional parameter to the Monitor.track_bunch_data method.
ProfileMonitor: if the bunch is empty, no profile is saved.
WakePotentialMonitor: if the bunch is empty, the same from last good turn is saved.
BunchSpectrumMonitor: if the bunch is empty, the incoherent spectrum is nan.

libver='earliest' allow for better backward compatibility when switching HDF5 versions.

Convert h5py data into numpy arrays as fancy indexing is broken in h5py >= 3.1.0

Check that the bunch is not empty before binning in Mpi.share_distributions. If it is then the beam filling pattern is updated and empty array are sent.
In CavityResonator.track, the tracking without mpi is modified to take into account the case with an empty bunch. Mpi case is fine as the beam filling pattern is updated in the call for Mpi.share_distributions.
In WakePotential.track, if the bunch is empty tracking does nothing.

Fix bug with chromaticity where chromaticity was taken as chromaticity * tune.

Fixed error on the computation of the beam size with dispersion in Synchrotron.sigma method.
Add import of plot_beamspectrum and streak_beamspectrum

Fix the sign of the dipolar wake potential to get the correct sign of the tune shift in tracking with respect to the impedance/wake function conventions used in mbtack2.

Import at inside the Optics.load_from_AT method to allow use of mbtrack2 without at installed.
Update load_from_AT

Fix BeamSpectrumMonitor
Add plotting functions for BeamSpectrumMonitor

Add plot_bunchspectrum and streak_bunchspectrum
Fix bug (double plot) in plot_bunchdata
Remove plot_tunedata

Remove TuneMonitor (remplaced by BunchSpectrumMonitor).
Remove default values for save_very, buffer_size, total_size.
Set default value for mpi_mode to False.
Add "xtau"/"ytau" choices for Bunch/BeamSpectrumMonitor classes.
Set default value for n_fft to None to avoid zero-padding by default.
Fix WakePotentialMontior n_bin (x2) problem.

Fix for BeamSpectrumMonitor and BunchSpectrumMonitor with merge_files

Add a monitor for the coherent beam spectrum (i.e. coupled bunch modes).
Add utility functions to BunchSpectrumMonitor to know the fft and signal resolution.
Save the frequency samples for BunchSpectrumMonitor in the hdf5 file.

Allow to monitor coherent and incoherent bunch spectrums.

Rework plot_beamdata and plot_bunchdata to be able to plot multiple hdf5 files at the same time
New function streak_beamdata to separate the streak plot option.

Beam.mpi_share_distributions is now moved to mpi.share_distribution
Now mpi4py is only needed for the multi-bunch tracking and not for single bunch

bunch_index returns the array with the positions (index) of the non-empty bunches, to be used with Beam.not_empty iterator.
bunch_mean, bunch_std and bunch_emit now iterate only over non empty bunches.

CavityResonator.init_phasor is now CavityResonator.init_phasor_track
New CavityResonator.init_phasor uses analytic formula
Fix phase slipping factor for initialization case (rf frame should be delta = (self.wr - self.m*self.ring.omega1))

fix init

Fix mpi_share_distributions

This avoid to store a huge amount of data in all cores before calling mpi_init to share it to all the cores.