Fix Async problem for exponential retry, clean up tests, and fix forking

[mjsteger]

Fixes problem noted in #100, and adds a test for the same. Cleans up the tests and splits them, allowing all the unit tests to be run without credentials(this is for circleci/local testing). Finally, fixing forking by removing the logging directly from the signal handlers, see here for more info on the issue. This was to allow more consistent running of the unit tests.

N.B. that the large number of additions/removals due to moving around/chunking the integration tests, as I was trying out a solution of running those on circle as well (but could not, likely due to some assumptions in the code that cause them to not run in parallel correctly)

[pdrakeweb]

In testing this PR, I was unable to use error_handler based asynchronous activity error handling to capture the error raised by asynchronous activities that use the exponential retry mechanism. Is that expected, or perhaps I was going about it improperly?

[mjsteger]

Uh, I'm not sure precisely what you mean. Could you provide an example? It's not expected that you cannot use the error_handler to handle errors in async activities with exponential retry, though it should be noted that they will get retried before they throw out.

[pdrakeweb]

I was finding that rather than the error_handler rescue block capturing the raised failure, it was causing the workflow to fail. It may be that I was doing something wrong, but I will try to reproduce the problem with a simple workflow and a couple activities that I can share.