[llvm] Fix loop cache cost to avoid cost of zero for refgroups. (PR #88915)

[via llvm-commits]

================
@@ -8,6 +8,9 @@
 ; Check IndexedReference::computeRefCost can handle type differences between
 ; Stride and TripCount
 
+; Round costs up to the nearest whole number i.e. in 'for.cond5' cost is calculated 12.5 and
----------------
RouzbehPaktinat wrote:

@nikic 
I changed it to use getUDivCeil  for all values as you suggested and made required changes to the impacted tests similar to this one.

https://github.com/llvm/llvm-project/pull/88915