[ET-VK][testing] Create dedicated test binary for pointwise convolutions

Pull Request resolved: #17220

This commit creates a dedicated test binary for pointwise (1x1) convolutions
(test_q8ta_conv2d_pw), separating them from the general 2D convolution tests.
Here are the key changes:

What Changed

1. New Test Binary: test_q8ta_conv2d_pw.cpp (591 lines)

   - Dedicated test file focusing exclusively on pointwise convolutions (kernel
     size 1x1)
   - Contains 9 test configurations ranging from accuracy tests to performance
     cases:
     - Accuracy tests: Various channel configurations (32→3, 64→32, 96→64, 13→7,
       80→40) with different spatial dimensions
     - Performance tests: Larger configurations (160→480, 22→48, 48→48, 128→128)
       exceeding the 100-dim reference limit
   - Tests all combinations of:
     - Storage types: Texture3D, Buffer
     - Int8 memory layouts: 4C1W, 4W4C, 4C
   - Also tests legacy 4W4C implementation via impl_selector="legacy_4w4c"
   - Includes full reference implementation for numerical correctness
     verification
   - Custom FLOP calculator for performance measurements

2. Removed from test_q8ta_conv2d.cpp (44 lines deleted)

   - Removed 6 pointwise convolution configurations that are now covered by the
     new dedicated binary
   - General conv2d tests now focus solely on kernels > 1x1 (3x3, 5x5, etc.)

3. Build System Updates

   - Added test_q8ta_conv2d_pw target to:
     - targets.bzl (Buck2)
     - CMakeLists.txt (CMake)
   - Both fbcode and xplat paths updated (files are mirrored)

4. CI/Workflow Integration

   - Updated executorch_vulkan_eureka_unit_tests.sky to include the new test
     binary in on-device testing workflow

Why This Separation?

Pointwise convolutions (1x1 kernels) are a distinct optimization target with
different performance characteristics than general convolutions. Separating them
enables:

- Focused performance iteration on pointwise-specific shaders
- Cleaner test organization
- Faster test runs when only testing one convolution type
ghstack-source-id: 338638548
@exported-using-ghexport

Differential Revision: [D92307251](https://our.internmc.facebook.com/intern/diff/D92307251/)

Author: ssjia

backends/vulkan/test/custom_ops/CMakeLists.txt

@@ -100,6 +100,7 @@ if(TARGET vulkan_backend)
   add_operator_prototype(test_q8ta_qdq)
   add_operator_prototype(test_q8ta_clone)
   add_operator_prototype(test_q8ta_conv2d)
+  add_operator_prototype(test_q8ta_conv2d_pw)
   add_operator_prototype(test_q8ta_conv2d_dw)
   add_operator_prototype(q8ta_q8ta_q8to_add)
 endif()

backends/vulkan/test/custom_ops/targets.bzl

@@ -94,5 +94,6 @@ def define_common_targets(is_fbcode = False):
     define_custom_op_test_binary("test_q8ta_qdq")
     define_custom_op_test_binary("test_q8ta_clone")
     define_custom_op_test_binary("test_q8ta_conv2d")
+    define_custom_op_test_binary("test_q8ta_conv2d_pw")
     define_custom_op_test_binary("test_q8ta_conv2d_dw")
     define_custom_op_test_binary("q8ta_q8ta_q8to_add")

backends/vulkan/test/custom_ops/test_q8ta_conv2d.cpp

@@ -258,35 +258,6 @@ static std::vector<TestCase> generate_quantized_conv2d_test_cases() {
   }
 
   std::vector<Conv2dConfig> configs = {
-      // Pointwise convolutions: kernel size 1x1
-      {OutInChannels(32, 3),
-       InputSize2D(64, 64),
-       KernelSize(1, 1),
-       Stride(1, 1),
-       Padding(0, 0),
-       Dilation(1, 1),
-       1},
-      {OutInChannels(64, 32),
-       InputSize2D(32, 32),
-       KernelSize(1, 1),
-       Stride(1, 1),
-       Padding(0, 0),
-       Dilation(1, 1),
-       1},
-      {OutInChannels(96, 64),
-       InputSize2D(16, 16),
-       KernelSize(1, 1),
-       Stride(1, 1),
-       Padding(0, 0),
-       Dilation(1, 1),
-       1},
-      {OutInChannels(13, 7),
-       InputSize2D(57, 33),
-       KernelSize(1, 1),
-       Stride(1, 1),
-       Padding(0, 0),
-       Dilation(1, 1),
-       1},
       // General 2D convolutions
       {OutInChannels(32, 3),
        InputSize2D(64, 64),
@@ -352,21 +323,6 @@ static std::vector<TestCase> generate_quantized_conv2d_test_cases() {
        Padding(2, 2),
        Dilation(1, 1),
        4},
-      // Performance cases (pointwise - will use im2col)
-      {OutInChannels(128, 128),
-       InputSize2D(128, 128),
-       KernelSize(1, 1),
-       Stride(1, 1),
-       Padding(0, 0),
-       Dilation(1, 1),
-       1},
-      {OutInChannels(128, 128),
-       InputSize2D(128, 128),
-       KernelSize(1, 1),
-       Stride(1, 1),
-       Padding(0, 0),
-       Dilation(1, 1),
-       1},
       // Performance cases (3x3 convs - will use im2col)
       {OutInChannels(32, 3),
        InputSize2D(256, 256),