fix: Avoid overflow when computing FoR's difference of min and max (#2549)

Author: robert3005

vortex-btrblocks/src/integer/stats.rs

@@ -73,10 +73,11 @@ impl ErasedStats {
             ErasedStats::U16(x) => (x.max - x.min) as u64,
             ErasedStats::U32(x) => (x.max - x.min) as u64,
             ErasedStats::U64(x) => x.max - x.min,
-            ErasedStats::I8(x) => (x.max - x.min) as u64,
-            ErasedStats::I16(x) => (x.max - x.min) as u64,
-            ErasedStats::I32(x) => (x.max - x.min) as u64,
-            ErasedStats::I64(x) => (x.max - x.min) as u64,
+            ErasedStats::I8(x) => (x.max as i16 - x.min as i16) as u64,
+            ErasedStats::I16(x) => (x.max as i32 - x.min as i32) as u64,
+            ErasedStats::I32(x) => (x.max as i64 - x.min as i64) as u64,
+            ErasedStats::I64(x) => u64::try_from(x.max as i128 - x.min as i128)
+                .vortex_expect("max minus min result bigger than u64"),
         }
     }