fix: use non-empty block in benchmarking to include outhash (#15715)

## What
Benchmarks now use non‑empty blocks with non‑zero outHashes.

## Why
Storing outHash == 0 costs only 2100 gas, while "real" (non‑zero) hashes
cost 22100 gas (I assume that in mainnet we will always have at least
one outbox message per L2 block). Our previous gas measurement was
therefore off by ~20 k gas per block.

## Testing

Before:
```
| propose                             | 390446          | 401059 | 400462 | 416897 | 100     |
```

After:
```
| propose                             | 411330          | 421890 | 421316 | 437781 | 100     |
```

Co-authored-by: Maddiaa <[email protected]>

Author: linoscope

l1-contracts/test/benchmark/happy.t.sol

@@ -124,7 +124,7 @@ contract BenchmarkRollupTest is FeeModelTestPoints, DecoderBase {
     CommitteeAttestation[] attestations;
   }
 
-  DecoderBase.Full full = load("empty_block_1");
+  DecoderBase.Full full = load("single_tx_block_1");
 
   uint256 internal constant SLOT_DURATION = 36;
   uint256 internal constant EPOCH_DURATION = 32;