fix(e2e): make PII test model name configurable via environment variable

The PII detection E2E test was hardcoding "model": "general-expert" in
the chat completions request, which caused failures in deployments that
use different model names (e.g., aibrix uses "vllm-llama3-8b-instruct").

This commit makes the model name configurable via the E2E_TEST_MODEL
environment variable, while maintaining backward compatibility by
defaulting to "general-expert" for existing profiles (ai-gateway,
dynamic-config).

Changes:
- Add E2E_TEST_MODEL environment variable support in pii_detection.go
- Default to "general-expert" when E2E_TEST_MODEL is not set
- Enable future profiles (e.g., aibrix) to specify custom model names

Testing:
- Tested locally with dynamic-config profile: 100/100 tests passed (100% accuracy)
- Tested locally with ai-gateway profile: 100/100 tests passed (100% accuracy)

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

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Yossi Ovadia <yovadia@redhat.com>

Author: yossiovadia

e2e/testcases/pii_detection.go

@@ -133,9 +133,15 @@ func testSinglePIIDetection(ctx context.Context, testCase PIITestCase, localPort
 		ExpectedBlocked: testCase.ExpectedBlocked,
 	}
 
+	// Get model name from environment, default to "general-expert" for backward compatibility
+	modelName := os.Getenv("E2E_TEST_MODEL")
+	if modelName == "" {
+		modelName = "general-expert"
+	}
+
 	// Create chat completion request
 	requestBody := map[string]interface{}{
-		"model": "general-expert",
+		"model": modelName,
 		"messages": []map[string]string{
 			{"role": "user", "content": testCase.Question},
 		},