Makes jsonrpc_handler more flexible

[RobRuana]

The jsonrpc_handler was created with the intent that it would be added as a member of a class, so it was given a single positional parameter named self. This unfortunately breaks when the jsonrpc_handler is used outside the context of a class (it throws "missing required self argument" errors).

The self parameter is never actually referenced. This pull request switches the parameter from self to self=None, so the jsonrpc_handler function can be invoked outside the context of a class.

[EliAndrewC]

+1

It looks like the Travis tests are failing for this branch but passed the last time they were run several months ago. However, it looks like they exit entirely as soon as the server tests begin, which leads me to believe that the problem isn't with this code but rather because Travis has perhaps begin to be more strict about whether we're allowed to do stuff like have a socket open a TCP port and make connections to it.

This leads me to believe that we could get our tests running again by changing this line
https://github.com/magfest/sideboard/blob/master/tox.ini#L32
to have -m 'not functional' and thus omit the "functional" tests which actually start and stop CherryPy from being run by Travis. This isn't ideal, but would probably be a good thing to try in the meantime.

If you feel like making and testing that change as part of this PR so that merging this doesn't break Travis then that would be great. If you don't want to hold up this PR since we have other stuff which depends on it, then I'd be okay with this getting merged in as-is and punting on getting Travis running again.