Improve key-rotation mechanism

[sideninja]

If there is a bigger load on transaction submission it might come to a case where there are less keys than transactions being submitted which might mean that the same key will be used before the previous transaction is sealed.
Improvement should be made to wait for that transaction to be sealed before submitting a new one. There's also a possibility with offline nonce tracking with a frequent sync.

[m-Peter]

Related to #477

[j1010001]

instead of sealed we could use just Tx executed

[m-Peter]

Completed by #693