Changes

| Mochitest failure investigator

| We spent countless hours investigating tests and managing known failures that occur occasionally. In many cases we have investigated test failures and cannot reproduce them while running the test by itself. It has been found that some tests alter the state of Firefox which can cause future tests to behave differently. Some tests are even written to depend on this state change.

To build this tool we need to take a given test failure and work backwards to find its failure point. Starting with the given test failure, we need to start a fresh instance of the browser/profile and verify the single test case passes on it's own. Then we can repeat this process by adding more tests for each run until we find the failure. The most logical way to add tests will be per directory as each directory usually contains tests of similiar behavior. This would be added into the mochitest python harness. Doing so will give people the ability to quickly run this locally and with a small adjustment to our Try server (passing in the correct command line arguments), we can run this on production machines.