Xtrack

**Changes:**
- Introdude multi-frame mode in ParticlesMonitor
- Added examples for ParticlesMonitor usage

**Fix:**
- Particles of species different from the reference particle correctly propagated in LossLocationRefinement

**Changes:**
- Global aperture check can be disabled passing Tracker(..., global_xy_limit=None).
- Particles lost on the global aperture limit are identified by -1 in the particle state.
- Fixes in LossLocationRefinement: correct propagation of global aperture settings, correct charge and mass in refinement (and in BeamInteraction. Global aperture check is disabled for backtracker.

**Changes:**
- Changed stored energy variable from psigma to ptau
- Cleaned up examples for mixed cpu/gpu simulations and for ring with synchrotron radiation
- MAD-X loader uses angle variable instead of knl[0] to set reference system curvature (used in tapered lattices)
- Dispersion calculation accounts for energy loss along the beam line (needed in presence of strong radiation)

**Changes:**
- Fix in import RF cavity from MAD-X when the harmonic number is provided.
- Fix in wire import from MAD-X.
- Example with Wakefield on CPU together with lattice tracking and spacecharge on GPU.
- Introduced test for mad import for several elements.
- Other minor fixes.

**Changes:**
- Custom damping rate and equilibrium emittance in LinearTransferMatrix
- Tracker can transfer the particles to the CPU context for specific elements exposing .needs_cpu flag.
- Tracker can hide hidden particles for specific elements exposing .needs_hidden_lost_particles flag.
- Tracker in collective mode cleans pyheadtail slices (if present) for each line part.
- Spacecharge installation in line is now done in place.
- Introduced s_tol flag in Line.insert_element
- Fix bugs in Wire.to_dict, Elens.to_dict, DipoleEdge.to_dict.
- Introduced behaves_like_drift flag for thick elements (allows twiss inside the element)
- Fix in matrix responsiveness check in compute_linear_normal_form
- Regenerated all json files in test_data.