v1.1.0

• Added f-DPG implemented by the FDPGTuner and FCDPGTueners. These tuners allow to tune models using other f-divergences, such as total variation (TVLoss), reverse KL (ReverseKLLoss), and Jensen-Shannon (JSLoss).
• Added estimates of acceptance rate and divergence to the target distribution to QRS
• LMDistribution tweaks:
  • switched to a model parameter & removed pad_token_id overwriting for seq2seq models
  • set seq2seq models to use empty context & tweaked scoring
  • switched from using nature to auto & improved handling of bos_token
  • added clone() method
  • fixed the timing of the feature moments reporting