Fix type used in bytes_hashed addition

`length` was changed to `bytes_hashed` and the type changed from `usize`
to `u64`.

The addition under the `hashes_fuzz` feature flag has been changed from
`usize` to `u64` to fix the fuzz error.

hashes/src/util.rs

@@ -91,7 +91,7 @@ macro_rules! engine_input_impl(
             for c in inp {
                 self.buffer[0] ^= *c;
             }
-            self.bytes_hashed += inp.len();
+            self.bytes_hashed += inp.len() as u64;
         }
     )
 );