Startup/shutdown leaks threads from build task queue executors

When refactoring the purge module, which brought more PlexusTestCase tests into continuum-core, running a clean build with:

mvn clean install

Started taking extremely long during continuum-core's test phase. Using jconsole I was able to see that the problem was that the JVM was managing an extremely large number of threads: approximately 450 when I connected. This ends up consuming a large amount of memory and bogs down the host system.

By cleanly disposing of task queue executors during application context shutdown, the will likely achieve better cpu and memory efficiency and stability.