feat: more efficient compute_l2_to_l1_hash (#11036)

This avoids unnecessary tracking of the length of the `BoundedVec`.

noir-projects/noir-protocol-circuits/crates/types/src/hash.nr

@@ -138,19 +138,19 @@ pub fn compute_l2_to_l1_hash(
     rollup_version_id: Field,
     chain_id: Field,
 ) -> Field {
-    let mut bytes: BoundedVec<u8, 160> = BoundedVec::new();
+    let mut bytes: [u8; 160] = std::mem::zeroed();
 
     let inputs =
         [contract_address.to_field(), rollup_version_id, recipient.to_field(), chain_id, content];
-    for i in 0..inputs.len() {
+    for i in 0..5 {
         // TODO are bytes be in fr.to_buffer() ?
         let item_bytes: [u8; 32] = inputs[i].to_be_bytes();
         for j in 0..32 {
-            bytes.push(item_bytes[j]);
+            bytes[32 * i + j] = item_bytes[j];
         }
     }
 
-    sha256_to_field(bytes.storage())
+    sha256_to_field(bytes)
 }
 
 pub fn silo_l2_to_l1_message(