Some fixes to lo-byte parser

- The lo-byte parser would incorrectly call buf.copy() with an offset at
  the end of the target buffer and a length of 0. Though this is
  theoretically a no-op, Buffer.copy() is doing bounds-checking on the
  source before taking the length into account.
- Clear bufs[] and bufBytes when consuming partial buffer.

lib/websocket.js

@@ -243,7 +243,17 @@ var WebSocket = function(url, proto) {
 
                 assert.equal(mbOff, bufsBytes);
 
-                buf.copy(mb, mbOff, off, i);
+                // Don't call Buffer.copy() if we're coping 0 bytes. Rather
+                // than being a no-op, this will trigger a range violation on
+                // the destination.
+                if (i > 0) {
+                    buf.copy(mb, mbOff, off, i);
+                }
+
+                // We consumed all of the buffers that we'd been saving; clear
+                // things out
+                bufs = [];
+                bufsBytes = 0;
             }
 
             process.nextTick(function() {