[FuncAttrs] Always preserve FunctionAnalysisManagerCGSCCProxy

[aeubanks]

Summary

FunctionAnalysisManagerCGSCCProxy should not be preserved if any of its
keys may be invalid. Since we are not removing/adding functions in
FuncAttrs, it's fine to preserve it.

[aeubanks]

This shouldn't help compile times yet since we still end up invalidating all the same analyses in FunctionAnalysisManagerCGSCCProxy::Result::invalidate, but it's a step in that direction.
I want to look into preserving analyses for functions not modified in a CGSCC pass, even if other functions are modified, to help D98103 and compile times in general.

[asbirlea]

Sounds good, thank you!

llvm/lib/Transforms/IPO/FunctionAttrs.cpp

auto AARGetter = [&](Function &F) -> AAResults & {
     return FAM.getResult<AAManager>(F);
   };
 
   SmallVector<Function *, 8> Functions;
   for (LazyCallGraph::Node &N : C) {
     Functions.push_back(&N.getFunction());
   }
 
-  if (deriveAttrsInPostOrder(Functions, AARGetter))
-    return PreservedAnalyses::none();
+  if (deriveAttrsInPostOrder(Functions, AARGetter)) {
+    // We have not changed the call graph or removed/added functions.
+    PreservedAnalyses PA;
+    PA.preserve<FunctionAnalysisManagerCGSCCProxy>();
+    return PA;
+  }
 
   return PreservedAnalyses::all();
 }
 
 namespace {
 
 struct PostOrderFunctionAttrsLegacyPass : public CallGraphSCCPass {
   // Pass identification, replacement for typeid