Add comment linking to closed PR for future optimizers

While optimizing these operations proved unfruitful w.r.t. improving
compiler performance right now, faster versions might be needed at a
later time.

compiler/rustc_index/src/bit_set.rs

@@ -654,15 +654,19 @@ impl<T: Idx> BitRelations<ChunkedBitSet<T>> for ChunkedBitSet<T> {
 
 impl<T: Idx> BitRelations<HybridBitSet<T>> for ChunkedBitSet<T> {
     fn union(&mut self, other: &HybridBitSet<T>) -> bool {
-        // FIXME: this is slow if `other` is dense, and could easily be
-        // improved, but it hasn't been a problem in practice so far.
+        // FIXME: This is slow if `other` is dense, but it hasn't been a problem
+        // in practice so far.
+        // If a a faster implementation of this operation is required, consider
+        // reopening https://github.com/rust-lang/rust/pull/94625
         assert_eq!(self.domain_size, other.domain_size());
         sequential_update(|elem| self.insert(elem), other.iter())
     }
 
     fn subtract(&mut self, other: &HybridBitSet<T>) -> bool {
-        // FIXME: this is slow if `other` is dense, and could easily be
-        // improved, but it hasn't been a problem in practice so far.
+        // FIXME: This is slow if `other` is dense, but it hasn't been a problem
+        // in practice so far.
+        // If a a faster implementation of this operation is required, consider
+        // reopening https://github.com/rust-lang/rust/pull/94625
         assert_eq!(self.domain_size, other.domain_size());
         sequential_update(|elem| self.remove(elem), other.iter())
     }