Refactor calibration

[cliffckerr]

• Ensure ss.Calibration().calibrate() works as easily as possible
• Simplify build_sim
• Update API to ensure consistency
• Improve plotting (e.g. include data)
• Simplify suggest_type
• Reconsider names for beta/beta_binomial etc
• Improve docstrings
• Consider weighted interval score / energy scores

Consider moving this to check_fit():

    if calib.after_fits.mean() <= calib.before_fits.mean():
        print('✓ Calibration improved fit')
    else:
        print('✗ Calibration did not improve fit, but this sometimes happens stochastically and is not necessarily an error')