fix: Avoid huge compilation times in base rollup (#9113)

This temporarily disables the checking of decomposition against P, that
together with the new sha256 noir implementation generates a blowup in
compilation time and ram usage. This is fine because we are going to
replace this with poseidon for blobs anyway

noir-projects/noir-protocol-circuits/crates/rollup-lib/src/components.nr

@@ -312,7 +312,8 @@ pub fn compute_tx_effects_hash(
 
     let mut hash_input_flattened = [0; TX_EFFECTS_HASH_INPUT_FIELDS * 32];
     for offset in 0..TX_EFFECTS_HASH_INPUT_FIELDS {
-        let input_as_bytes: [u8; 32] = tx_effects_hash_input[offset].to_be_bytes();
+        // TODO: This is not checking that the decomposition is smaller than P
+        let input_as_bytes: [u8; 32] = tx_effects_hash_input[offset].to_be_radix(256);
         for byte_index in 0..32 {
             hash_input_flattened[offset * 32 + byte_index] = input_as_bytes[byte_index];
         }

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

@@ -212,7 +212,8 @@ pub fn compute_tx_logs_hash(logs: [LogHash; MAX_ENCRYPTED_LOGS_PER_TX]) -> Field
     // Convert each field element into a byte array and append the bytes to `hash_input_flattened`
     let mut hash_input_flattened = [0; MAX_ENCRYPTED_LOGS_PER_TX * 32];
     for offset in 0..MAX_ENCRYPTED_LOGS_PER_TX {
-        let input_as_bytes: [u8; 32] = logs[offset].value.to_be_bytes();
+        // TODO: This is not checking that the decomposition is smaller than P
+        let input_as_bytes: [u8; 32] = logs[offset].value.to_be_radix(256);
         for byte_index in 0..32 {
             hash_input_flattened[offset * 32 + byte_index] = input_as_bytes[byte_index];
         }
@@ -233,7 +234,8 @@ pub fn compute_tx_note_logs_hash(logs: [LogHash; MAX_NOTE_ENCRYPTED_LOGS_PER_TX]
     // Convert each field element into a byte array and append the bytes to `hash_input_flattened`
     let mut hash_input_flattened = [0; MAX_NOTE_ENCRYPTED_LOGS_PER_TX * 32];
     for offset in 0..MAX_NOTE_ENCRYPTED_LOGS_PER_TX {
-        let input_as_bytes: [u8; 32] = logs[offset].value.to_be_bytes();
+        // TODO: This is not checking that the decomposition is smaller than P
+        let input_as_bytes: [u8; 32] = logs[offset].value.to_be_radix(256);
         for byte_index in 0..32 {
             hash_input_flattened[offset * 32 + byte_index] = input_as_bytes[byte_index];
         }