[llvm] [Float2Int] Fix miscompile with floats that can be converted to large values (PR #85996)

[Craig Topper via llvm-commits]

================
@@ -349,3 +349,29 @@ bogusBB:                                          ; preds = %bogusBB
   %tobool = fcmp une double %inc, 0.000000e+00
   br label %bogusBB
 }
+
+define i32 @pr71958() {
----------------
topperc wrote:

This should be 79158

https://github.com/llvm/llvm-project/pull/85996