More graceful stack overflow handling

[kleinj]

As pointed out in #77, if a stack overflow occurs in one of the computations threads (e.g., SCC analysis in a "deep" model, very heavily nested formulas/expressions, ...), the GUI does not notice that the computation thread is gone as the error handling there only catches Exceptions and not Errors. Errors are generally serious and should not be caught, but for a StackOverflowError we can gracefully recover by just ending the computation thread, the same as for other exceptions.

This PR addresses:

• Making the stack size configurable (via -javastack x command-line parameter)
• Gracefully handling stack overflows in the GUI (logging short stack trace to the log, short message with hint about -javastack in the pop-up) and command-line (short stack trace in the log & hint).
• If the stack overflow arises in the GUI during a model checking computation, it is recorded as the error in the result object, like other errors.

Have not tested on Linux / Windows yet, but don't expect trouble...

For the GUI, I think I covered the most likely situations (steady-state computation, verification, experiments) in testing. The test case for #77 is good for triggering the overflow in the explicit engine, either for steady-state computation or using a S=?[ ... ] property.