fix(aibrix): revert PII model to ModernBERT configuration from PR #688

The aibrix E2E profile was designed to use ModernBERT for PII detection
(as configured in PR #688), but during the merge it was incorrectly
changed to use the LoRA PII model which isn't available in the aibrix
deployment.

This commit reverts the aibrix PII model configuration to use:
- model_id: "models/pii_classifier_modernbert-base_presidio_token_model"
- use_modernbert: true

This restores the original working configuration from PR #688 and allows
the aibrix E2E tests to pass while keeping the LoRA PII auto-detection
feature available for other profiles that explicitly configure it.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>
Signed-off-by: Yossi Ovadia <[email protected]>

Author: yossiovadia

deploy/kubernetes/aibrix/semantic-router-values/values.yaml

@@ -438,13 +438,11 @@ config:
       use_cpu: true
       category_mapping_path: "models/category_classifier_modernbert-base_model/category_mapping.json"
     pii_model:
-      # Support both traditional (modernbert) and LoRA-based PII detection
-      # When model_type is "auto", the system will auto-detect LoRA configuration
-      model_id: "models/lora_pii_detector_bert-base-uncased_model"
-      model_type: "auto"  # Enables LoRA auto-detection
+      model_id: "models/pii_classifier_modernbert-base_presidio_token_model"
+      use_modernbert: true
       threshold: 0.7
       use_cpu: true
-      pii_mapping_path: "models/lora_pii_detector_bert-base-uncased_model/pii_type_mapping.json"
+      pii_mapping_path: "models/pii_classifier_modernbert-base_presidio_token_model/pii_type_mapping.json"
 
   keyword_rules:
     - category: "thinking"