Outstanding long polling requests need to be aborted when the pool/assignment is disabled

[cdhowie]

If the pool or pool-worker assignment corresponding to an outstanding long polling request is disabled, the long polling request does not abort. This has three implications:

• The worker will likely not make a new long polling request after the next getwork and the worker will therefore not receive long polling notifications from the new active pool. (If the new active pool does not support long polling requests, this is not an issue.)
• When the outstanding long polling request completes, the worker will begin working on that work, which will be form a disabled pool. Submitted shares will still be routed to the pool, but the worker shouldn't be working for a disabled pool.
• If the worker gets the URI for the next long polling request from the response for the first long polling request, the worker will continue to poll the wrong pool indefinitely.