fix(CRITICAL): CoreXGDN interface mismatch + engine death protection

Three fixes for the three bugs in latest docker log:

1. corex_gdn.py REWRITTEN — interface now matches qwen3_5.py:
   OLD: CoreXGDN(num_heads, head_dim, layer_idx, chunk_size, eps)
   NEW: CoreXGDN(num_v_heads, num_k_heads, head_k_dim, head_v_dim, conv_kernel_size, layer_idx)

   OLD forward: (q, k, v, gate, beta, conv_state, temporal_state, attn_metadata)
   NEW forward: (hidden_states, attn_metadata, conv_state, temporal_state,
                  in_proj_qkv, in_proj_z, in_proj_b, in_proj_a,
                  conv1d_weight, A_log, dt_bias, norm, out_proj)

   Fixes: 'CoreXGDN.__init__() got unexpected keyword argument num_v_heads'

2. serving_chat.py — engine death protection for multimodal:
   When model has no multimodal_config, return 400 instead of passing image data
   to engine (which causes permanent AsyncEngineDeadError).

   Fixes: 'ValueError: You set image=0 but found 1 items'

3. patch_ops.sh — ALWAYS deploy our modules (base image has bugs):
   - qwen3_5.py: ALWAYS deploy (base has NaN)
   - corex_gdn/moe/fa2.py: ALWAYS deploy (base interface mismatch)
   - corex_fa2.py was MISSING from base → now deployed

qwen3_6_scripts/serving_chat.py

@@ -312,6 +312,14 @@ class OpenAIServingChat(OpenAIServing):
             engine_inputs = TokensPrompt(
                 prompt_token_ids=prompt_inputs["prompt_token_ids"])
             if mm_data is not None:
+                # Protect engine from death: if model doesn't support multimodal,
+                # return 400 instead of crashing the entire engine.
+                # ValueError "image=0 but found 1" kills the async engine permanently.
+                mm_config = getattr(self.model_config, 'multimodal_config', None)
+                if mm_config is None:
+                    logger.warning("Image data in request but model has no multimodal_config — rejecting to protect engine")
+                    return self.create_error_response(
+                        "This model does not support multimodal (image) inputs.")
                 engine_inputs["multi_modal_data"] = mm_data
 
             is_tracing_enabled = (await