clarify casting of tmp

Author: jayz22

soroban-sdk/src/crypto/bls12_381.rs

@@ -25,6 +25,8 @@ pub struct Bls12_381 {
 // https://github.com/arkworks-rs/algebra/blob/bf1c9b22b30325ef4df4f701dedcb6dea904c587/ff/src/biginteger/arithmetic.rs#L66-L79
 fn sbb_for_sub_with_borrow(a: &mut u64, b: u64, borrow: u8) -> u8 {
     let tmp = (1u128 << 64) + u128::from(*a) - u128::from(b) - u128::from(borrow);
+    // casting is safe here because `tmp` can only exceed u64 by a single
+    // (borrow) bit, which we capture in the next line.
     *a = tmp as u64;
     u8::from(tmp >> 64 == 0)
 }