[LoopVectorize] Revert Classic Flang patch causing AArch64 test failure

This patch partially reverts the following commit from an early version of
Classic Flang LLVM:

flang-compiler/llvm@6866909

When the LoopVectorize.cpp change in that commit is ported to LLVM 14, it causes
a failure in the following test:

llvm/test/Transforms/LoopVectorize/AArch64/smallest-and-widest-types.ll

Reverting the change allows all tests to run cleanly.

Author: bryanpkc

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

@@ -4816,10 +4816,8 @@ void LoopVectorizationCostModel::collectElementTypesForWidening() {
       if (ValuesToIgnore.count(&I))
         continue;
 
-      // Examine Loads, Stores, PHINodes
-      // Also examine instructions which convert to a float/double
-      if (!isa<LoadInst>(I) && !isa<StoreInst>(I) && !isa<PHINode>(I) &&
-          !isa<FPExtInst>(I) && !isa<SIToFPInst>(I) && !isa<UIToFPInst>(I))
+      // Only examine Loads, Stores and PHINodes.
+      if (!isa<LoadInst>(I) && !isa<StoreInst>(I) && !isa<PHINode>(I))
         continue;
 
       // Examine PHI nodes that are reduction variables. Update the type to