Update on "[ET-VK] Contiguous buffer implementation for embedding"

Title says it all!

Also moved around some structs throughout the `glslh` files for better code organization.

Differential Revision: [D84716456](https://our.internmc.facebook.com/intern/diff/D84716456/)

[ghstack-poisoned]

Author: ssjia

backends/vulkan/test/tester.py

@@ -44,8 +44,9 @@ def __init__(
 
 class Partition(BaseStages.Partition):
     def __init__(self, partitioner: Optional[Partitioner] = None):
+        vk_compile_spec = {"skip_bool_tensors": True}
         super().__init__(
-            partitioner=partitioner or VulkanPartitioner(),
+            partitioner=partitioner or VulkanPartitioner(vk_compile_spec),
         )
 
 
@@ -55,6 +56,10 @@ def __init__(
         partitioners: Optional[List[Partitioner]] = None,
         edge_compile_config: Optional[EdgeCompileConfig] = None,
     ):
+        if partitioners is None:
+            vk_compile_spec = {"skip_bool_tensors": True}
+            partitioners = [VulkanPartitioner(vk_compile_spec)]
+
         super().__init__(
             default_partitioner_cls=VulkanPartitioner,
             partitioners=partitioners,