Claim is too slow while "processDepositStatus" handle too much deposits

[chengzhinei]

There is a serious problem when monitorTxs handle "ReviewMonitoredTx" and update the mtx.Nonce:

here, the mTx.Nonce may bigger than nonce, and the nonce will never updated. And all the mTx can not be sended.
We have saw this bug happend in X1, and it made the bridge function hanged until we use this fixed code.
And this bug caused to the next problem, the GER include more than 70000 txs, and it took more than 10 hours to create claimTx.
A ExitRoots may include many deposits, and during "processDepositStatus()", it builds claimTx(one claimTx may cost about 1s), but only when storage.Commit, the ClaimTx will be write into db, the whole process may cost more than 10 minutes. But during that time, the "monitorTxs" can't get the claimTxs, and no claimTx would be sended to L2, so the claim is slowed down.
And I made a change, every deposit builds a claimTx, it stores the claimTx into db immediately.

The changes code only increases the speed of writing ClaimTx into storage, the logic of sending ClaimTx is not influenced.

PR:#544