[CSSPGO] Turn on postlink sample counts overwrite.

[hoy]

Summary

The CS preinliner computes sample counts for outlined functions globally, resulting in a more accurate function profile than the actual thinlto postlink sample inliner does. I've seen that turning on --overwrite-existing-weights for postlink results in general 0.5% to 1% code size win (perf wise is netural) for CSSPGO accross four of our big internal services, but slight code size regression for non-CS AutoFDO. I'm turning on the switch for preinlined profile only.

llvm/lib/Transforms/IPO/SampleProfile.cpp

if (!CallsitePrioritizedInline.getNumOccurrences())
       CallsitePrioritizedInline = true;
     // For CSSPGO, we also allow recursive inline to best use context profile.
     if (!AllowRecursiveInline.getNumOccurrences())
       AllowRecursiveInline = true;
 
     if (Reader->profileIsPreInlined()) {
       if (!UsePreInlinerDecision.getNumOccurrences())
         UsePreInlinerDecision = true;
+      if (!OverwriteExistingWeights.getNumOccurrences())
+        OverwriteExistingWeights = true;
     }
 
     if (!Reader->profileIsCS()) {
       // Non-CS profile should be fine without a function size budget for the
       // inliner since the contexts in the profile are either all from inlining
       // in the prevoius build or pre-computed by the preinliner with a size
       // cap, thus they are bounded.
       if (!ProfileInlineLimitMin.getNumOccurrences())