LLM REPL with shell_gpt

This PR implements a REPL for a large language model (in this case implemented by [shell_gpt](https://github.com/TheR1D/shell_gpt). As part of this interface, I also wrote a code that can yank the output of the last command using `CTRL-Y` for the ipython and sgpt terminals, which partially resolves #10. @GoldsteinE do you have any comments?