Implement end points for native commands

[vijayvammi]

It is better to have a method in place to allow users to run commands natively instead of Popen, especially as a pytorch has issues with it.

One way to allow it is to allow users to write a script, roughly along the lines of which becomes the docker CMD.

#! /bin/bash
runnable pre-command -c config -p parameters -t tag ---run-id <>

runnable post-command -c config -p parameters -t tag ---run-id <>

The pre-command and post-command use local executor and have run log, catalog and secrets functionality.

Should be relatively easy to implement and an easy win.