[Feature] Support to use fullgraph with eagle (#5118)

### What this PR does / why we need it?
    
We support to use full graph with eagle. 

Change list:
1. Distinguish between processing graph_params and draft_graph_params in
attention_v1.
    2. Adapt the full-graph mode in eagle_proposer, include:
        1). If use full graph, make Fullgraph Wrapper when load model.
2). Build a new meatadata, set running mode in FULL and mark attention
update in dummy_run when in Fullgraph mode.
3). Fixed and fill any attn_metadata, such as
attn_metadata.slot_mapping.
        4). Add a descriptor.
        5). Set running mode and triggered update metadata.
3. Trans is_mtp_model to is_draft_model, and add the update of
workspace.

NOTE:
When set async_scheduling=True, the draft model will enforce execution
in eager mode.

### Does this PR introduce _any_ user-facing change?

### How was this patch tested?

- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c

---------

Signed-off-by: anon189Ty <Stari_Falcon@outlook.com>
Co-authored-by: Yizhou Liu <liu_yizhou@outlook.com>
Co-authored-by: Yizhou <136800916+yiz-liu@users.noreply.github.com>

vllm_ascend/compilation/acl_graph.py

@@ -254,7 +254,10 @@ def _update_attn_pa_params(update_stream, forward_context, runtime_shape):
 
 
 def _update_attn_fia_params(update_stream, forward_context, runtime_shape):
-    graph_params = get_graph_params()
+    if forward_context.is_draft_model:
+        graph_params = get_draft_graph_params()
+    else:
+        graph_params = get_graph_params()
     # For Qwen3-next, since the kv_cache_config has already categorized
     # linear_attn and self_attn, the attn_metadata is first arranged with
     # self_attn followed by linear_attn. Therefore, using zip directly
@@ -306,8 +309,8 @@ def update_attn_params(update_stream, forward_context, runtime_shape,
 
 def update_mla_attn_params(update_stream, forward_context, runtime_shape,
                            speculative_config):
-    if forward_context.is_mtp_model:
-        graph_params = get_mtp_graph_params()
+    if forward_context.is_draft_model:
+        graph_params = get_draft_graph_params()
     else:
         graph_params = get_graph_params()
     # FIXME: Behold! We are using a temporary hack here to update the args
@@ -326,7 +329,7 @@ def update_mla_attn_params(update_stream, forward_context, runtime_shape,
             seq_lens_list = forward_context.attn_metadata[
                 key].decode.seq_lens_list
             if speculative_config and speculative_config.method == "mtp" \
-                    and not forward_context.is_mtp_model:
+                    and not forward_context.is_draft_model:
                 actual_seq_lengths = forward_context.attn_metadata[
                     key].decode.actual_seq_lengths_q
                 spec_multiple = speculative_config.num_speculative_tokens + 1
@@ -336,7 +339,7 @@ def update_mla_attn_params(update_stream, forward_context, runtime_shape,
                     spec_multiple * (i + 1)
                     for i in range(runtime_shape // spec_multiple)
                 ]
-            elif forward_context.is_mtp_model:
+            elif forward_context.is_draft_model:
                 actual_seq_lengths = forward_context.attn_metadata[
                     key].decode.actual_seq_lengths_q
                 block_table = forward_context.attn_metadata[
@@ -440,8 +443,8 @@ def update_attn_dcp_pcp_params(update_stream, forward_context, runtime_shape):
 
 def update_mla_attn_dcp_pcp_params(update_stream, forward_context,
                                    runtime_shape):
-    if forward_context.is_mtp_model:
-        graph_params = get_mtp_graph_params()
+    if forward_context.is_draft_model:
+        graph_params = get_draft_graph_params()
     else:
         graph_params = get_graph_params()
     # FIXME: Behold! We are using a temporary hack here to update the args
@@ -527,14 +530,14 @@ def get_graph_params():
     return _graph_params
 
 
-_mtp_graph_params: Optional[GraphParams] = None
+_draft_graph_params: Optional[GraphParams] = None
 
 
-def set_mtp_graph_params(aclgraph_capture_sizes: list[int]):
-    global _mtp_graph_params
-    if _mtp_graph_params is not None:
-        raise ValueError("MTPGraph parameters have already been set!")
-    _mtp_graph_params = GraphParams(
+def set_draft_graph_params(aclgraph_capture_sizes: list[int]):
+    global _draft_graph_params
+    if _draft_graph_params is not None:
+        raise ValueError("DraftGraph parameters have already been set!")
+    _draft_graph_params = GraphParams(
         {size: []
          for size in aclgraph_capture_sizes},
         {size: None
@@ -546,11 +549,11 @@ def set_mtp_graph_params(aclgraph_capture_sizes: list[int]):
     )
 
 
-def update_mtp_graph_params_workspaces(num_tokens: int, workspace: Any):
-    global _mtp_graph_params
-    if _mtp_graph_params is not None:
-        _mtp_graph_params.workspaces[num_tokens] = workspace
+def update_draft_graph_params_workspaces(num_tokens: int, workspace: Any):
+    global _draft_graph_params
+    if _draft_graph_params is not None:
+        _draft_graph_params.workspaces[num_tokens] = workspace
 
 
-def get_mtp_graph_params():
-    return _mtp_graph_params
+def get_draft_graph_params():
+    return _draft_graph_params