fix rustdoc caching tests

[syphar]

After #1569 the tests failed, first because of 2db2d76, then because of something else.

I don't know why this is breaking, didn't have more time to dig this evening, but this fixes the test.

cc @jsha

[jsha]

Sorry about the breakage! This change looks good, though I am similarly unsure why they would cause a test to pass where it didn't before.

[Nemo157]

I did a little more investigation. Previously the test was passing after 5 seconds because iron would timeout the connection for the first response body and allow the second response to go through. In #1814 I added a 2 second timeout on the client-side to avoid 90s long test failures when idle connections exhaust the server, but that meant the client gave up on the second request before the server would timeout the first request and start processing it.

This must just be the only test that attempts to do two requests without either reading the body or dropping the first request (adding a resp.text().unwrap() in between the tests similarly fixes it since reqwest will then close the connection after reading the body).