Dispatching of isolated/device analysis jobs can fill work pool

[JonatanWaern]

• Needs to investigate whether this is a 'natural' result of jobs having long runtimes or if jobs hang
• Would be alleviated by outdated in-progress job cancelling
• Consider sending a message to user in these cases, warning that the DLS is overly busy

[JonatanWaern]

Clarification: it is not analysis jobs themselves that fill the work pool, but rather requests to the server while its under load