Support Bamboo LM sparse and dense inference (#171)

* support inference for sparse mistral arch

* patch convert script for the new model type

* fix: runtime constants for new model kind

* renaming: bamboo arch

* fix: model type

* bugfix: wrong arch when converting bamboo HF models

* chore: remove debugging info

* add script for dense model conversion (for Bamboo LM)

* support dense inference of Bamboo LM

* Add news in README.md

* Add readme for BambooLM and dense inference

* update todos