Fix file path in tiered prefix guide (llm-d#1785)

Signed-off-by: Cong Liu <conliu@google.com>

Author: liu-cong

guides/tiered-prefix-cache/modelserver/gpu/vllm/native/fs/gke/kustomization.yaml

@@ -1,7 +1,7 @@
 apiVersion: kustomize.config.k8s.io/v1beta1
 kind: Kustomization
 resources:
-  - ..
+  - ../base
 components:
   - ../../../../../../../recipes/modelserver/components/disable-gke-nccl-tuner-patch
 patches:

guides/tiered-prefix-cache/router/tiered-prefix-cache-cpu.values.yaml

@@ -18,13 +18,10 @@ router:
         - type: no-hit-lru-scorer
         - type: approx-prefix-cache-producer
           name: gpu-prefix-cache-producer
-          parameters:
-            maxPrefixTokensToMatch: 40000 # enough to cover the ~20k input length in the benchmark.
         - type: approx-prefix-cache-producer
           name: cpu-prefix-cache-producer
           parameters:
             autoTune: false
-            maxPrefixTokensToMatch: 40000 # enough to cover the ~20k input length in the benchmark.
             lruCapacityPerServer: 82000  # Allocating ~200GB for Qwen-32B: 82,000 blocks * 2.5MB/block (based on 160KB/token * 16 block size).
         - type: queue-scorer
         - type: kv-cache-utilization-scorer