use SIGIO as a way to detect stopped connections

[Andrew Dalke <dalke@…>]

Discussion at the  Sprint pointed out that we should have a way for potentially long-running jobs to detect if the http connection was closed, so they could abort what they are doing.

(It might be doing a complex calculation, or waiting for some other HTTP server or database server.)

Dave B proposed adding support for the SIGIO exception, and Andrew suggested making that available in a context manager, for use by those long-running handlers. (This detection will not be done automatically. Handler code must enable it on their own.)