Properly handle future blocks in PoA service

[netrome]

Follow-up from #2145 (comment)

Description

Currently, the extract_block_info function in the PoA service will return the current instant for last_block_created when the last block is in the future. We should either error on this case or find a good representation for this future instant, and properly handle this case in the rest of the PoA service.

Definition of done

The handling of last blocks with timestamps higher than the current system time is clear, correct and well-tested.

[netrome]

As part of this follow-up we could consider extending the time interface if helpful as discussed in #2145 (comment)

[MitchTurner]

We can't think of a specific use case for this immediately. We are going to keep this around because it's an interesting problem, but keep it relatively low priority.