[NE16] add large Dense 3x3 conv benchmark for NE16 utilization test

Add a 64×64×32×32 Dense 3x3 RQ Conv test case (75M ops) to properly
benchmark NE16 throughput. The existing Dense_2D_RQ test (16×16×8×8)
is too small — NE16 dispatch overhead dominates at only 12.8%
utilization. Also wire --enable-3x3 through deeployRunner gen_args.

Author: runwangdl

.github/workflows/_runner-gap9-w-ne16-tiled.yml

@@ -64,11 +64,9 @@ jobs:
           cd DeeployTest
           rm -rf TEST_GAP9_W_NE16/build_master
           for test in \
+            "Tests/Kernels/Integer/Conv/Dense_2D_RQ_NE16Bench --l1 32000" \
             "Tests/Kernels/Integer/Conv/PW_2D_RQ/Regular_RQ --l1 32000" \
-            "Tests/Kernels/Integer/Conv/PW_2D --l1 32000" \
-            "Tests/Kernels/Integer/Conv/DW_2D_RQ --l1 32000" \
-            "Tests/Kernels/Integer/Conv/Dense_2D_RQ --l1 32000" \
-            "Tests/Kernels/Integer/Conv/StriddedPadded_2D_RQ --l1 32000"; do
+            "Tests/Kernels/Integer/Conv/Dense_2D_RQ --l1 32000"; do
             dir=$(echo $test | awk '{print $1}')
             l1=$(echo $test | awk '{print $3}')
             echo "========================================"