test: Ephemeral tree hashPreimage

[IlyasRidhuan]

The benchmark run actually also doubles as a useful test vector since it updates a "pre-filled" leaf.

This test is also useful in that it shows that we don't need the condition on the hashPreimage . The only time the getKey() === 0n condition is used in the worldstate (which wasnt right in the ephemeral trees anyways) is when we are using the forced null inclusion on public data tree updates - which the avm trees handle separately (and isn't long for this world anyways).

tl;dr
Old version

hashPreimage<T extends TreeLeafPreimage>(preimage: T): Fr {
      // Watch for this edge-case, we are hashing the key=0 leaf to 0.
      // This is for backward compatibility with the world state implementation
      if (preimage.getKey() === 0n) {
        return Fr.zero();
      }
      const input = preimage.toHashInputs().map(x => Fr.fromBuffer(x));
      return poseidon2Hash(input);
    }

New Version

hashPreimage<T extends TreeLeafPreimage>(preimage: T): Fr {
     const input = preimage.toHashInputs().map(x => Fr.fromBuffer(x));
     return poseidon2Hash(input);
   }

[dbanks12]

Approved, but it looks like it's failing 🤔
Edit: saw your msg after I reviewed. Makes sense. I'll make sure the fix works before merging this.