Avoid seg fault with recent nixpkgs

Resolves #16 (closed) .

I wouldn't say I understand the semantics of C++, but this seems to fix the seg fault. My very cursory understanding is that the previous code was wrong because resp would be deleted when it went out of scope, whereas now it doesn't as it lives on the heap instead. Though this might leak Response values.

What confuses me is why this behaviour changed at all...

stub/cbits/stub.cpp

@@ -208,8 +208,6 @@ class Response {
         trace("finishing with code: %d\n", status);
         this->flush(status);
     }
-
-    Response(Response &x) = delete;
 };
 
 static bool paused = false;
@@ -606,7 +604,7 @@ static int handle_command(Socket& sock, const char *buf, uint32_t cmd_len) {
 
 
       case CMD_POLL:
-        r_poll_pause_resp = &resp;
+        r_poll_pause_resp = new Response(resp);
         // NOTE: Don't call finish so that the process blocks waiting for
         // a response. We will send the response when the process pauses.
         break;