The maximum number of threads performing sequencing and text extraction operations should be configurable

When performing sequencing & text extraction operations, ModeShape uses background threads for executing the actual work. These threads come (at the moment) from an unbounded cached thread pool meaning that it's possible in some cases and on some OSes (Windows being a very likely culprit) for java.lang.OutOfMemoryError: unable to create new native thread to occur.

We should make sure that:
1. we're always using a default upper bound for the number of threads that can be spawned
2. this number is configurable so that clients can fine-tune it if required