Update VirtualSocketServerTest to use a fake clock.

Since this is a test for a fake network, it's only natural that it uses
a fake clock as well. This makes the tests much faster, less flaky, and
lets them be moved out of  "webrtc_nonparallel_tests", since they no
longer have a dependency on any "real" thing (sockets, or time) and
can be run in parallel as easily as any other tests.

As part of this CL, added the fake clock as an argument to
VirtualSocketServer's and TestClient's constructors, since these classes
have methods that wait synchronously for something to occur, and if the
test is using a fake clock, they need to advance it in order to make
progress.

Lastly, added a DCHECK in Thread::ProcessMessages. If called with a
nonzero time while a fake clock is used, it will get stuck in an
infinite loop; a DCHECK is easier to notice than an infinite loop.

BUG=webrtc:7727, webrtc:2409

Review-Url: https://codereview.webrtc.org/2927413002
Cr-Commit-Position: refs/heads/master@{#18544}

webrtc/base/thread.cc

@@ -471,6 +471,11 @@ void Thread::Clear(MessageHandler* phandler,
 // Note that these methods have a separate implementation for mac and ios
 // defined in webrtc/base/thread_darwin.mm.
 bool Thread::ProcessMessages(int cmsLoop) {
+  // Using ProcessMessages with a custom clock for testing and a time greater
+  // than 0 doesn't work, since it's not guaranteed to advance the custom
+  // clock's time, and may get stuck in an infinite loop.
+  RTC_DCHECK(GetClockForTesting() == nullptr || cmsLoop == 0 ||
+             cmsLoop == kForever);
   int64_t msEnd = (kForever == cmsLoop) ? 0 : TimeAfter(cmsLoop);
   int cmsNext = cmsLoop;