[BOLT][DWARF] Fix performance regression running BOLT on binaries build with DWARF4

[ayermolo]

Summary

In one of the previous diffs LocBuffer was changed to pass by value. This lead to
performance regression running BOLT on binaries with DWARF4 split dwarf.

[maksfb]

Looks good, but please rephrase the summary as it's unclear what "this was changed to pass by value" means when one is not looking at the code. Since it's a performance fix, remove "NFC" from the title.

bolt/lib/Core/DebugData.cpp

static void writeEmptyListDwarf5(raw_svector_ostream &Stream) {
   support::endian::write(
       Stream, static_cast<uint8_t>(dwarf::DW_LLE_end_of_list), support::little);
 }
 
 static void writeLegacyLocList(DIEValue &AttrInfo,
                                DebugLocationsVector &LocList,
                                DIEBuilder &DIEBldr, DIE &Die,
                                DebugAddrWriter &AddrWriter,
-                               DebugBufferVector LocBuffer, DWARFUnit &CU,
+                               DebugBufferVector &LocBuffer, DWARFUnit &CU,
                                raw_svector_ostream &LocStream) {
   if (LocList.empty()) {
     replaceLocValbyForm(DIEBldr, Die, AttrInfo, AttrInfo.getForm(),
                         DebugLocWriter::EmptyListOffset);
     return;
   }
 
   const uint32_t EntryOffset = LocBuffer.size();