chore: bind single_query_cached_kv_attention to python

kernels/attention.cpp

@@ -0,0 +1,19 @@
+#include <torch/extension.h>
+
+void single_query_cached_kv_attention(
+    torch::Tensor& out,
+    torch::Tensor& query,
+    torch::Tensor& key_cache,
+    torch::Tensor& value_cache,
+    float scale,
+    torch::Tensor& block_tables,
+    torch::Tensor& context_lens,
+    int block_size,
+    int max_context_len);
+
+PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
+    m.def(
+        "single_query_cached_kv_attention",
+        &single_query_cached_kv_attention,
+        "Compute the attention between an input query and the cached key/value tensors.");
+}