Add the ability to send Expect-Staple reports.

net/http/transport_security_state.h

Index: net/http/transport_security_state.h
diff --git a/net/http/transport_security_state.h b/net/http/transport_security_state.h
index d5283712c57ae28e53a6f63cecd722c136bc7a19..68dfc08ae91f2f8a097e06073c9ea709bb1c888e 100644
--- a/net/http/transport_security_state.h
+++ b/net/http/transport_security_state.h
@@ -449,6 +449,10 @@ class NET_EXPORT TransportSecurityState
                              const HostPortPair& host_port_pair,
                              const SSLInfo& ssl_info);
 
+  void ProcessExpectStaple(const HostPortPair& host_port_pair,
+                           const SSLInfo& ssl_info,
+                           const std::string& ocsp_response);

[Ryan Sleevi, 2016/07/19 00:02:49]

Documentation is needed here. Just from the header, it's unclear why
ocsp_response is needed.

From a design standpoint, the naming here leaves much to be desired - I thought
at first this processed ExpectStaple headers, but no, it performs the logic of
ExpectStaple processing.

From a documentation standpoint, it's unclear how one would use this. For
example, do you call this for all hosts when you get a stapled response? Do you
only call it if the host is in the expect-staple list? Do you only call it when
the stapled response is bad?

I would try, as much as possible, to examine the other methods of TSS and try to
make it consistent with them.

[dadrian, 2016/07/19 00:18:54]

I modeled this after ProcessExpectCTHeader. It follows almost identical
structure, modulo one less reporting abstract class. It doesn't take a header
because ExpectStaple is preload-only. The plan is to call this from
ssl_client_socket_impl.cc in DoVerifyCertComplete().

That being said, I totally forgot to put any documentation on this. :)

+
   // For unit tests only; causes ShouldRequireCT() to return |*required|
   // by default (that is, unless a RequireCTDelegate overrides). Set to
   // nullptr to reset.