semaphore.c: decrease count before release sem_lock

This fixes a hang in test_multiprocessing when running without the GIL
due to lost modifications to count in the seamphore.

Modules/_multiprocessing/semaphore.c

@@ -191,6 +191,7 @@ _multiprocessing_SemLock_release_impl(SemLockObject *self)
         assert(self->count == 1);
     }
 
+    --self->count;
     if (!ReleaseSemaphore(self->handle, 1, NULL)) {
         if (GetLastError() == ERROR_TOO_MANY_POSTS) {
             PyErr_SetString(PyExc_ValueError, "semaphore or lock "
@@ -201,7 +202,6 @@ _multiprocessing_SemLock_release_impl(SemLockObject *self)
         }
     }
 
-    --self->count;
     Py_RETURN_NONE;
 }
 
@@ -437,10 +437,10 @@ _multiprocessing_SemLock_release_impl(SemLockObject *self)
 #endif
     }
 
+    --self->count;
     if (sem_post(self->handle) < 0)
         return PyErr_SetFromErrno(PyExc_OSError);
 
-    --self->count;
     Py_RETURN_NONE;
 }