Don't stake for previously staked times

[alfiedotwtf]

Currently, each call to produce_block() will stake from the "last block timestamp" to 'last block timestamp + some chain configuration" value...

In the produce_block() call, if there is ever a recoverable error, we re-attempt to stake until we hit that maximum time limit. If still no successful block has been solved, we return with a TryAgainLater. This is fine for a single call to produce_block() as it prevents staking over timeframes that we have already attempted, but this currently doesn't prevent rework on subsequent calls.

This fix will make every call to produce_block() restrict the search space to only timestamps we have not yet searched.