Don't check for unflushed bytes in SSLClientSocketOpenSSL::IsConnectedAndIdle.

Don't check for unflushed bytes in SSLClientSocketOpenSSL::IsConnectedAndIdle.

This aligns with SSLClientSocketNSS's behavior (which never checks the ring
buffer) and with TCPClientSocket's behavior (which never checks writes at all).
Because SSLClientSocket::Write may complete before bytes are actually written
to the network, it's possible there is data in transport_bio_ still to be
written to the network.

Normally this wouldn't happen as the peer would read the request before sending
a response. But this breaks POSTing to a URL that requires HTTP Negotiate as
the server may return a response before draining the POST body. Due to the
delayed flush, transport_bio_ may not be clear by the time we check
IsConnectedAndIdle.

This is not the ideal fix to the auth issue, but it's a minimal one and aligns
our OpenSSL behavior with our old NSS one. IsConnectedAndIdle is somewhat
unavoidably weird and racy since the check it makes is inherently not very
sound.

BUG=466147
Committed: https://crrev.com/fc9a6b8c1e98be8ab7b3da6589770abc155a37d6
Cr-Commit-Position: refs/heads/master@{#325348}

[davidben, 2015-04-15 17:51:59]

davidben@chromium.org changed reviewers:
+ rsleevi@chromium.org

[davidben, 2015-04-15 17:51:59]

I'm not suuuuper happy with this change, but I think it's worthwhile. Both as a
minimal fix (revising HTTP auth code is risky and definitely not something to
merge to 43) and to align with SSLClientSocketNSS. It's kind of weird, but if
you look at all the other IsConnectedAndIdle checks, they're all really weird.

Really I'd like to change SSLClientSocketOpenSSL, possibly with a tweaked
BoringSSL API, so that we don't need to do this early-return anywhere. Transport
read and write errors should come in on the higher-level request which triggered
them.

[Ryan Sleevi, 2015-04-15 21:50:39]

lgtm

[davidben, 2015-04-15 22:10:32]

The CQ bit was checked by davidben@chromium.org

[commit-bot: I haz the power, 2015-04-15 22:10:53]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1089123003/20001

[commit-bot: I haz the power, 2015-04-15 23:47:42]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-04-15 23:50:00]

Patchset 2 (id:??) landed as
https://crrev.com/fc9a6b8c1e98be8ab7b3da6589770abc155a37d6
Cr-Commit-Position: refs/heads/master@{#325348}