Use unittest's built in test discovery method

[objarni]

If we could use unittests built in test discovery system, we could potentially get rid of a lot of code from pytddmon.

We have to make sure it acts just as pytddmon acts (e.g. all systests pass).

Note that this would not deprecate the file-monitoring functionality of pytddmon, however combined with the cheesechop and a low-level file system monitoring API, pytddmon could actually shrink dramatically in number of lines.

[objarni]

The command line, in Python 2.7, is this:
python -m unittest discover.

What needs investigation:

• command line for 2.6, 3.2 and 2.3

Is command line running/parsing viable? What fidelity do we loose? Can we use unittest module discover programmatically instead?

We still need process isolation (currently implemented by multiprocessing pool. What a hack!).

Relevant function - find_files

(and possibly run_tests, run_tests_in_file, run_module if we get test running "for free" too)

[objarni]

The big advantage of this is reducing number of lines of code in pytddmon.py, decreasing maintenance costs and improving quality.