There is no such concept in Tomcat/JBoss Web. All web apps run in the same JVM (hence in the same process) and there is no mechanism to provide priority or throttling of requests on a web app basis.
Request handling starts in the Connector and is configurable in server.xml. The Connector has properties e.g., port and thread pool metrics. If one webapp supports long-lived requests starving other requests from obtaining a thread, you can adjust the thread pool metrics to avoid starvation.
If you want to isolate the applications to avoid thread contention the best way is to deploy two distinct instances of JBossWeb. These days that's fairly easy to do using virtualization.
Another possibility - configure an additional HTTP Connector to listen on a unique port. The additional connector as its own thread pool by definition.