Do we need to separate the tensorflow_fs?

[bravomikekilo]

tensorflow_fs is a greatly reduced TensorFlow. It only hold the filesystem and logging, event, record reader and writer. Except power mxconsole as a native library, maybe it can be used in other project