A few small perf improvements

Plus a benchmark of the getattr() function.
I have some ideas on how to make this faster, but I don't
think it's that important at the moment.

microbenchmarks/getattrfunc_ubench.py

+class C(object):
+    pass
+
+def f():
+    g = getattr
+    c = C()
+    c.o = 1
+    for i in xrange(10000000):
+        g(c, "o")
+f()

src/runtime/dict.cpp

@@ -249,6 +249,8 @@ extern "C" PyObject* PyDict_GetItem(PyObject* dict, PyObject* key) noexcept {
         return d->getOrNull(key);
     }
 
+    // XXX this would be easy to make much faster.
+
     // This path doesn't exist in CPython; we have it to support extension modules that do
     // something along the lines of PyDict_GetItem(PyModule_GetDict()):
     try {
@@ -304,6 +306,9 @@ extern "C" int PyDict_Next(PyObject* op, Py_ssize_t* ppos, PyObject** pkey, PyOb
 }
 
 extern "C" PyObject* PyDict_GetItemString(PyObject* dict, const char* key) noexcept {
+    if (dict->cls == attrwrapper_cls)
+        return unwrapAttrWrapper(dict)->getattr(key);
+
     Box* key_s;
     try {
         key_s = boxString(key);

src/runtime/list.cpp

@@ -594,6 +594,10 @@ Box* listIAdd(BoxedList* self, Box* _rhs) {
 
         int s1 = self->size;
         int s2 = rhs->size;
+
+        if (s2 == 0)
+            return self;
+
         self->ensure(s1 + s2);
 
         memcpy(self->elts->elts + s1, rhs->elts->elts, sizeof(rhs->elts->elts[0]) * s2);