[ET-VK] Re-implement split_with_sizes (#15793)

Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at
bottom):
* #15829
* #15796
* #15795
* #15794
* __->__ #15793

As title. The current implementation of split_with_sizes uses functions
from the `Copy.[h|cpp]` file in particular
`add_copy_channel_offset_node`. However, the shaders dispatched by this
function have a critical bug where the output tensor is passed in
separately with difference access types, i.e.

```cpp
graph.execute_nodes().emplace_back(new DispatchNode(
        graph,
        VK_KERNEL_FROM_STR(kernel_name),
        global_size,
        local_size,
        // Inputs and Outputs
        {
            {out, vkapi::kWrite},
            {out, vkapi::kRead},
            {in, vkapi::kRead},
        },
```

This creates many validation layer errors because the memory barriers
for the resource cannot be formed properly. The shader essentially
relies on undefined behaviour to work correctly

To fix, this diff re-implements the operator from scratch with a
dedicated compute shader.

Differential Revision:
[D86910642](https://our.internmc.facebook.com/intern/diff/D86910642/)

---------

Co-authored-by: ssjia <[email protected]>

Author: SS-JIA

backends/vulkan/op_registry.py

@@ -740,6 +740,7 @@ def register_cat_op():
     [
         exir_ops.edge.aten.select_copy.int,
         exir_ops.edge.aten.slice_copy.Tensor,
+        exir_ops.edge.aten.split_with_sizes_copy.default,
     ]
 )
 def register_transfer_ops():
@@ -782,10 +783,7 @@ def register_ported_op():
 # Ops ported from PyTorch Vulkan backend. These ops are in a separate registry because they support all packed dimensions
 @update_features(
     [
-        # Tensor combination
         exir_ops.edge.aten.repeat.default,
-        exir_ops.edge.aten.split_with_sizes_copy.default,
-        exir_ops.edge.aten.split.Tensor,
     ]
 )
 def register_ported_op_all_packed_dims():

backends/vulkan/runtime/graph/ops/glsl/common.glslh

@@ -86,4 +86,15 @@ int quantize_and_pack(const vec4 vals, const float inv_scale, const int zp) {
   return pack_into_int32(quantized);
 }
 
+#ifdef DEBUG_MODE
+
+#define printf debugPrintfEXT
+
+void printVec4(vec4 texel) {
+  debugPrintfEXT(
+      "texel: %f, %f, %f, %f\\n", texel.x, texel.y, texel.z, texel.w);
+}
+
+#endif // DEBUG_MODE
+
 #endif // COMMON_GLSLH