Keep-Alive ignored within Service error handling.

[ben-crowhurst]

We don't check for 'Connection: keep-alive' before calling close on 404, 501, and 500 error logic.

[ben-crowhurst]

This can't be properly tested until we have a more flexible HTTP client, see issue #41.

[ben-crowhurst]

The default service/resource fault handler has been designed to perform as little work as possible. This decision was made to reduce the likelihood of causing additional error conditions and mudding up the water. Users are free to set their own error handlers via Service::set_error_handler/Resource::set_error_handler if they do not agree with this decision.