Add merged Falcon 7B model with LoRA weights and tokenizer, and README

initial commit