[Database] L1 message `L1MessageKey::BlockNumber` lookup edge case

[frisitano]

We should consider the case where the last message in a batch is skipped and update the L1MessageKey::BlockNumber query to account for it.

Originally posted by @frisitano in #380

In practice, this should never happen as we no longer skip L1 messages and we never had a skipped L1 message at the end of a batch historically. However, in the name of being comprehensive it would be good to conver this edge case.

[yiweichi]

Close this by PR #402