Stop checking in VmapGeneratedPlumbing.h (pytorch#82351)

This PR changes VmapGeneratedPlumbing.h to be generated by torchgen. The
output file is ATen/VmapGeneratedPlumbing.h.

Why generate this file inside PyTorch codegen instead of a separate step
in functorch?
- I can't figure out how to get functorch's fbcode target to generate
- functorch's build system will, in the mid-term, be absorbed into
pytorch's build system, so I don't want to do the extra work of adding
a step to the functorch build process.

Test Plan:
- build pytorch, build functorch
Pull Request resolved: pytorch#82351
Approved by: https://github.com/ezyang

BUILD.bazel

@@ -79,6 +79,7 @@ generated_cpu_cpp = [
     "aten/src/ATen/MethodOperators.h",
     "aten/src/ATen/NativeMetaFunctions.h",
     "aten/src/ATen/RegistrationDeclarations.h",
+    "aten/src/ATen/VmapGeneratedPlumbing.h",
     "aten/src/ATen/core/aten_interned_strings.h",
     "aten/src/ATen/core/enum_tag.h",
     "aten/src/ATen/core/TensorBody.h",

build.bzl

@@ -144,6 +144,7 @@ GENERATED_H = [
     "FunctionalInverses.h",
     "RedispatchFunctions.h",
     "RegistrationDeclarations.h",
+    "VmapGeneratedPlumbing.h",
 ]
 
 GENERATED_H_CORE = [

functorch/functorch/csrc/BatchRulesHelper.h

@@ -17,7 +17,7 @@
 #include <functorch/csrc/PlumbingHelper.h>
 #include <ATen/core/dispatch/Dispatcher.h>
 #include <functorch/csrc/Constants.h>
-#include <functorch/csrc/VmapGeneratedPlumbing.h>
+#include <ATen/VmapGeneratedPlumbing.h>
 
 namespace at { namespace functorch {
 Tensor reshape_dim_into(int64_t src, int64_t dst, const Tensor& x);

functorch/functorch/csrc/BatchingMetaprogramming.h

@@ -6,7 +6,7 @@
 
 #pragma once
 #include <ATen/Tensor.h>
-#include <functorch/csrc/VmapGeneratedPlumbing.h>
+#include <ATen/VmapGeneratedPlumbing.h>
 
 namespace at {
 namespace functorch {