Modify discrete log to support sync and async - #405

It seems like overkill to go from global state to state within a class instance, but for most common usages of discrete_log, there won't be a practical difference.

The decision to raise an exception is a bit more interesting, because now you need to audit all the places that ask for a discrete log and make sure that they're prepared to catch that exception or it's okay for them to fail.

In my version, I just had it return Optional[int], which caused more pain initially, but the payoff was that mypy helped me find each and every place that was depending on that value succeeding, and then add more error handling where needed.

BTW, as to having the maximum value be 1 million or 100 million, you want this value to be big enough to encompass the homomorphic tally of the biggest election that ElectionGuard will ever handle. In L.A. County, California (which is probably the biggest single election jurisdiction in the U.S.), there were somewhere north of 4 million votes cast in November 2020. So, I picked 100 million to be really, really bigger than that, but still small enough that if you computed the whole thing by accident, you wouldn't run for a huge amount of time only to fail.

Also, if you really want to go the o-o route, I'd kill off the top-level functions and inline them into the DiscreteLog class, since anybody outside of this file shouldn't have to know about which mutex or which cache to use.