I simplified the logic in pythonexc2js and implemented better error handling. I'm pretty sure that the original version actually always leaks the exception state because PyErr_Fetch takes ownership of type, value, and traceback but it never decref'd them.

Another issue with the original code is that it calls PyErr_Print() when the error indicator is unset. The docs say that PyErr_Print() may cause a fatal system error if called when the error indicator is unset:
https://docs.python.org/3/c-api/exceptions.html?highlight=pyerr_print#c.PyErr_PrintEx
This is not apparently true, looking at the source code of PyErr_Print, there is a guard check whether the error indicator is set, if not just does nothing. (This is really what it should do, not like it costs much to be reasonable.) Anyways, because the documented behavior of PyErr_Print says it triggers a fatal error in this case, I figure we ought to try to avoid calling it in that condition.

In future PRs I am hoping to update the error formatting to alternate between Python and javascript stack traces as the code crosses between languages and to implement fatal error trapping.