Better defaults for production (in particular adaptive time step)

[MaxThevenet]

• Turn on predict and phase options of adaptive time step by default. They make a huge difference in the presence of density ramps, see below.
• Turn on beam sub cycling by default
• Decrease nt_per_betatron default. Should be hardly less accurate and 2x faster
• Fix a few print statements.

The figure below shows the final emittance after a PWFA stage, as a function of nt_per_betatron (the number of time steps per betatron period with adaptive time step), with the original adaptive time step, with feature predict on, with feature phase on, and with both. It clearly converges much faster with both. The right plot shows the effect of the tolerance of the phase feature.
adaptive_new.pdf

[MaxThevenet]

CI fails probably because sub-cycling is now on by default. I could reset the 3 failing tests.

[MaxThevenet]

After some more test, I observed little difference with the subcycling. We could put this up to 5, but that can be done in another PR.

[SeverinDiederichs]

Great! Thank you for the detailed study and drastically improving the adaptive time step 🚀