PR Contains:

1. Preliminary solution to Issue Connect to an already running Matlab instance instead of making one #196.
2. Reorganization of code to facilitate Matlab scripts that use pymatbridge to call Python functions. Separation of concerns (moving _preamble_code to matlabserver helper functions) is how I think the solution to clear potential hang conditions #13 should look.
3. Replaced all "print" and *.write() statements with configurable logging.
4. Tried to update code comments that were out of date.

The new convention for connecting to Matlab is that

lab = Matlab(executable=something, ...)

follows the old behavior, while

lab = Matlab(executable=None, socket_addr='ipc:///something', ...)

connects to a Matlab sessions where the matlabserver.m function is being executed.

The only thing I'm having trouble getting to work is the new "separate" command, designed for when you've started a Matlab session via Matlab(executable=matlab).start() and later decide that what you need to do requires (or would just be nicer) if you were to drop into Matlab proper without having to save, exit, and reload the workspace. Combined with the ability to dynamically connect to Matlab, this could facilitate workflows that involve actively switching back and forth.

On my box, calling something like
lab = Matlab()
lab.start()
lab.separate()
opens the Matlab desktop GUI, as expected, but the ZMQ socket and context closing messages are mysteriously printed twice, and the GUI eventually exists without becoming available for use. Any ideas here?