[MC] Speed up AttemptToFoldSymbolOffsetDifference in the absence of MCAsmLayout

The `FA < FB` check added by https://reviews.llvm.org/D153096 is slow.
Compute an informal layout order to speed up computation when
`AttemptToFoldSymbolOffsetDifference` is repeatedly called for the same
section.

Commit 9500a5d ("[MC] Make UseAssemblerInfoForParsing mostly true")
exposed this performance pitfall, which was mitigated by
`setUseAssemblerInfoForParsing(false)` workarounds (e.g. commit
245491a). The workaround can be removed
now.

Author: MaskRay

llvm/lib/MC/MCExpr.cpp

@@ -675,8 +675,14 @@ static void AttemptToFoldSymbolOffsetDifference(
     if (FA == FB) {
       Reverse = SA.getOffset() < SB.getOffset();
     } else if (!isa<MCDummyFragment>(FA)) {
-      Reverse = std::find_if(std::next(FA->getIterator()), SecA.end(),
-                             [&](auto &I) { return &I == FB; }) != SecA.end();
+      // Testing FA < FB is slow. Use setLayoutOrder to speed up computation.
+      // The formal layout order will be finalized in MCAssembler::layout.
+      if (FA->getLayoutOrder() == 0 || FB->getLayoutOrder()== 0) {
+        unsigned LayoutOrder = 0;
+        for (MCFragment &F : *FA->getParent())
+          F.setLayoutOrder(++LayoutOrder);
+      }
+      Reverse = FA->getLayoutOrder() < FB->getLayoutOrder();
     }
 
     uint64_t SAOffset = SA.getOffset(), SBOffset = SB.getOffset();