[Hybrid] Added supports_mamba_prefix_caching Protocol (#27339)

Signed-off-by: asafg <39553475+Josephasafg@users.noreply.github.com>

vllm/model_executor/models/bamba.py

@@ -37,7 +37,14 @@ from vllm.model_executor.layers.vocab_parallel_embedding import (
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 from vllm.sequence import IntermediateTensors
 
-from .interfaces import HasInnerState, IsHybrid, SupportsLoRA, SupportsPP, SupportsQuant
+from .interfaces import (
+    HasInnerState,
+    IsHybrid,
+    SupportsLoRA,
+    SupportsMambaPrefixCaching,
+    SupportsPP,
+    SupportsQuant,
+)
 from .utils import (
     AutoWeightsLoader,
     is_pp_missing_parameter,
@@ -394,7 +401,13 @@ class BambaModel(nn.Module):
 
 
 class BambaForCausalLM(
-    nn.Module, HasInnerState, SupportsLoRA, SupportsPP, IsHybrid, SupportsQuant
+    nn.Module,
+    HasInnerState,
+    SupportsLoRA,
+    SupportsPP,
+    IsHybrid,
+    SupportsQuant,
+    SupportsMambaPrefixCaching,
 ):
     packed_modules_mapping = {
         "qkv_proj": [