Update on "[ET-VK] double, short, and uint16 dtype runtime support"

Creating support for double, short, and uint16 for quantization ops. Registering the short keyword since theres already support.

Also changing the cpu implementation to support half

Differential Revision: [D75959063](https://our.internmc.facebook.com/intern/diff/D75959063/)

[ghstack-poisoned]

Author: morelos

backends/vulkan/test/op_tests/utils/gen_benchmark_vk.py

@@ -177,6 +177,8 @@ def generate_benchmark_fixture(self) -> str:
 
 vkapi::ScalarType from_at_scalartype(c10::ScalarType at_scalartype) {{
   switch (at_scalartype) {{
+    case c10::kDouble:
+      return vkapi::kDouble;
     case c10::kFloat:
       return vkapi::kFloat;
     case c10::kHalf:
@@ -187,6 +189,8 @@ def generate_benchmark_fixture(self) -> str:
       return vkapi::kInt;
     case c10::kChar:
       return vkapi::kChar;
+    case c10::kBool:
+      return vkapi::kBool;
     default:
       VK_THROW("Unsupported at::ScalarType!");
   }}

backends/vulkan/test/op_tests/utils/gen_correctness_vk.py

@@ -119,15 +119,11 @@ def gen_parameterization(self) -> str:
     case c10::kInt:
       return vkapi::kInt;
     case c10::kLong:
-      return vkapi::kLong;
-    case c10::kShort:
-      return vkapi::kShort;
+      return vkapi::kInt;
     case c10::kChar:
       return vkapi::kChar;
     case c10::kBool:
       return vkapi::kBool;
-    case c10::kByte:
-      return vkapi::kByte;
     default:
       VK_THROW("Unsupported at::ScalarType!");
   }