Fix the type of paged_kv_cache in append (#597)

The type is adjusted to be consistent with the prefill/decode wrappers.

python/flashinfer/page.py

@@ -197,7 +197,7 @@ def append_paged_kv_cache(
     append_value: torch.Tensor,
     batch_indices: torch.Tensor,
     positions: torch.Tensor,
-    paged_kv_cache: torch.Tensor,
+    paged_kv_cache: Union[torch.Tensor, Tuple[torch.Tensor, torch.Tensor]],
     kv_indices: torch.Tensor,
     kv_indptr: torch.Tensor,
     kv_last_page_len: torch.Tensor,