[Speculative Decoding] Unify Spec and non-spec branch (#6685)

* optimize spec-inference architecture

* delete debug log

* optimize spec_method usage  && fix unit_test

* add claude unit-test skill

* fix some ugly bug

* enhance robustness and bounds check

* unify method & spec_method to method to avoid bug

* activate CI

* fix unit test

* Unify logprobs computation for naive and speculative decoding, fix CUDA kernel

* fix logprob bug && optimize verify kernel

* fix exist_decode() judge

fastdeploy/model_executor/graph_optimization/cudagraph_piecewise_backend.py

@@ -159,7 +159,7 @@ class CudaGraphPiecewiseBackend:
         real_shape = ids_remove_padding.shape[0]
         if self.speculative_decoding and all(self.real_bsz_to_captured_size.values()):
             seq_lens_this_time: paddle.Tensor = kwargs["forward_meta"].seq_lens_this_time
-            num_running_requests = seq_lens_this_time.flatten().nonzero(as_tuple=False)[-1].item() + 1
+            num_running_requests = int((seq_lens_this_time.flatten() > 0).sum().item())
             real_shape = self.real_bsz_to_captured_size[num_running_requests]
         exist_prefill = kwargs["forward_meta"].exist_prefill
         # Static split graph mode: use Static + CUDAGraph for prefill/mixed phase