Depend on extension/threadpool, not thread_parallel_interface, in buck (pytorch#9511)

I planned to do this everywhere and forgot. Clean it all up, leave a
note, enforce the note with visibility. This makes sure everything in
buck-land gets ET_USE_THREADPOOL.

Test Plan: Profiled run on internal model, no longer seeing
parallel_for_no_threadpool

Author: swolchok

extension/parallel/targets.bzl

@@ -17,6 +17,6 @@ def define_common_targets():
             "@EXECUTORCH_CLIENTS",
         ],
         deps = [
-            "//executorch/runtime/kernel:thread_parallel_interface",
+            "//executorch/extension/threadpool:threadpool",
         ],
     )

kernels/optimized/cpu/targets.bzl

@@ -99,8 +99,8 @@ _OPTIMIZED_ATEN_OPS = (
     op_target(
         name = "op_where",
         deps = [
+            "//executorch/extension/threadpool:threadpool",
             "//executorch/kernels/portable/cpu/util:elementwise_util",
-            "//executorch/runtime/kernel:thread_parallel_interface",
         ],
     ),
 )

kernels/optimized/lib_defs.bzl

@@ -232,9 +232,9 @@ def define_libs(is_fbcode=False):
                 "DEFAULT": [],
             }) + LIBBLAS_DEPS,
             exported_deps = [
+                "//executorch/extension/threadpool:threadpool",
                 "//executorch/kernels/optimized:libutils",
                 "//executorch/runtime/core/exec_aten:lib",
-                "//executorch/runtime/kernel:thread_parallel_interface",
             ],
             **get_apple_framework_deps_kwargs(is_fbcode),
         )

kernels/portable/cpu/util/targets.bzl

@@ -12,6 +12,7 @@ def define_common_targets():
     runtime.cxx_library(
         name = "all_deps",
         deps = [
+            "//executorch/extension/threadpool:threadpool",
             "//executorch/kernels/portable/cpu/util:functional_util",
             "//executorch/kernels/portable/cpu/util:broadcast_util",
             "//executorch/kernels/portable/cpu/util:kernel_ops_util",
@@ -32,7 +33,6 @@ def define_common_targets():
             "//executorch/kernels/portable/cpu/util:slice_util",
             "//executorch/kernels/portable/cpu/util:elementwise_util",
             "//executorch/kernels/portable/cpu/util:upsample_util",
-            "//executorch/runtime/kernel:thread_parallel_interface",
         ],
         visibility = ["//executorch/...", "@EXECUTORCH_CLIENTS"],
     )
@@ -111,7 +111,7 @@ def define_common_targets():
             ":broadcast_util",
             ":dtype_util",
             "//executorch/runtime/kernel:kernel_runtime_context",
-            "//executorch/runtime/kernel:thread_parallel_interface",
+            "//executorch/extension/threadpool:threadpool",
         ],
         deps = [
             "//executorch/kernels/portable/cpu:scalar_utils",
@@ -245,7 +245,7 @@ def define_common_targets():
         srcs = [],
         exported_headers = ["functional_util.h"],
         exported_deps = [
-            "//executorch/runtime/kernel:thread_parallel_interface",
+            "//executorch/extension/threadpool:threadpool",
         ],
         deps = [
             "//executorch/runtime/kernel:kernel_includes",
@@ -319,7 +319,7 @@ def define_common_targets():
                 "//executorch/runtime/core/exec_aten/util:tensor_util{}".format(suffix),
             ],
             exported_deps = [
-                "//executorch/runtime/kernel:thread_parallel_interface",
+                "//executorch/extension/threadpool:threadpool",
             ],
             exported_preprocessor_flags = ["-DUSE_ATEN_LIB"] if aten_mode else [],
             visibility = [

runtime/kernel/targets.bzl

@@ -59,9 +59,9 @@ def define_common_targets():
             "//executorch/runtime/core/portable_type/c10/c10:c10",
             "//executorch/runtime/platform:platform",
         ],
+        # Don't depend on this target, depend on //executorch/extension/threadpool:threadpool.
         visibility = [
-            "//executorch/...",
-            "@EXECUTORCH_CLIENTS",
+            "//executorch/extension/threadpool/...",
         ],
     )