Hardcoded end-to-end benchmark on MobileNet v1 in QS8 format

- Weights are random and quantization parameters are mocked
- In lieu of QS8 Global Average Pooling, QU8 version is used

PiperOrigin-RevId: 324936126
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 99e2730..362084e 100755
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -3712,6 +3712,7 @@
   ADD_LIBRARY(bench-models STATIC
     models/fp16-mobilenet-v1.cc
     models/fp32-mobilenet-v1.cc
+    models/qs8-mobilenet-v1.cc
     models/fp16-mobilenet-v2.cc
     models/fp32-mobilenet-v2.cc
     models/fp32-mobilenet-v3-large.cc