[turboprop] Don't use function context specialization with Turboprop

For turboprop, it's a better tradeoff to reuse the code than
specialising the code for a particular closure especially given we
optimize quite early when compared to Turbofan.

Bug: v8:9684
Change-Id: Icf5d8548bbdcac9e202dcf44c68e06cc4c732ba7
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2461242
Commit-Queue: Mythri Alle <mythria@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#70451}

src/compiler/pipeline.cc

@@ -1140,7 +1140,8 @@ PipelineCompilationJob::Status PipelineCompilationJob::PrepareJobImpl(
   if (compilation_info()->closure()->raw_feedback_cell().map() ==
           ReadOnlyRoots(isolate).one_closure_cell_map() &&
       !compilation_info()->is_osr() &&
-      !compilation_info()->IsNativeContextIndependent()) {
+      !compilation_info()->IsNativeContextIndependent() &&
+      !compilation_info()->IsTurboprop()) {
     compilation_info()->set_function_context_specializing();
     data_.ChooseSpecializationContext();
   }

test/mjsunit/mjsunit.status

@@ -1201,6 +1201,11 @@
   # lands
   'regress/regress-932953': [FAIL],
 
+  # Tests failing for the lack of function context specialization in Turboprop.
+  'compiler/abstract-equal-receiver': [FAIL],
+  'compiler/constant-fold-cow-array': [FAIL],
+  'compiler/promise-resolve-stable-maps': [FAIL],
+
   # https://crbug.com/v8/10894
   'math-floor-of-div': [SLOW],
 }],  # variant == turboprop