Fix detaching a debugger while threads are suspended.

The interesting part of this change is in "thread_list.cc".

I've done a TODO in TagFromClass, but haven't seen it make any practical
difference in a debugger. I also tightened up the types in GetThreadStatus
while investigating the fact that we report some threads as "RUNNING, SUSPENDED",
which makes no sense until you realize that TS_RUNNING corresponds to both
our kRunnable thread state and our kNative thread state, the latter of which
may actually be a suspended thread.

I've also made us fail faster in the "address in use" jdwp failure case,
and tidied up a bunch of the capitalization in logging.

Change-Id: I0fe705791d07db31c4615addce44da4fdfbfd0d1
diff --git a/src/jdwp/jdwp_handler.cc b/src/jdwp/jdwp_handler.cc
index b30a6a2..a11c805 100644
--- a/src/jdwp/jdwp_handler.cc
+++ b/src/jdwp/jdwp_handler.cc
@@ -925,13 +925,13 @@
 
   LOG(VERBOSE) << StringPrintf("  Req for status of thread 0x%llx", threadId);
 
-  uint32_t threadStatus;
-  uint32_t suspendStatus;
+  JDWP::JdwpThreadStatus threadStatus;
+  JDWP::JdwpSuspendStatus suspendStatus;
   if (!Dbg::GetThreadStatus(threadId, &threadStatus, &suspendStatus)) {
     return ERR_INVALID_THREAD;
   }
 
-  LOG(VERBOSE) << "    --> " << JdwpThreadStatus(threadStatus) << ", " << JdwpSuspendStatus(suspendStatus);
+  LOG(VERBOSE) << "    --> " << threadStatus << ", " << suspendStatus;
 
   expandBufAdd4BE(pReply, threadStatus);
   expandBufAdd4BE(pReply, suspendStatus);