Add timeout to NLP resolver

[joshua-stauffer]

When the NLP service receives a query to transform text, it first runs the text through a ML model, and then sends subqueries out to the Geo and ReadModel services in parallel. The issues is that if one of those services are down, a result will never return, and the NLP service will keep a reference to that task in memory indefinitely. This isn't an issue for the end user, since the API service will automatically timeout requests that take too long (currently 5 seconds). I'd like to add similar logic to the NLP Resolver class, and automatically remove queries that exceed a constant timeout.

Another factor is that the NLP processing itself can be very time consuming. A nice solution may be to refactor the text processing call from the base NLPService class into the Resolver class, and initialize a TTL value on creation. Then, after each processing step the TTL could be checked and the process aborted if the timeout has been exceeded.