Speculative block repeatedly restarted when near block boundary

[heifner]

When speculatively executing trxs if a block is exhausted then it is now immediately restarted via: #1319. However, when starting a speculative block we set a block deadline according to pending block time minus producer time offset. This can create very small block times. A trx then can exceed that block time but not the configured trx max time resulting in a burst of abort_block/start_block calls.

Example scenario:

• A speculative block is created with a time limit of pending block time minus producer offset. This can be rather small depending on when a block was received and how much time was spent processing the block
• trx exceeds block time and is rescheduled for the next block
• abort block
• new block created with the same time limit as the one just aborted
• trx exceeds block time and is rescheduled for the next block
• abort block
• repeat until block interval is exceeded or an incoming block is applied

When not producing a block, the time limit of a block should be set to fc::time_point::maximum() since we want to allow as many speculative trxs as possible until another block is received. If the consensus max cpu or max net is exceeded then the block will be restarted as the trx/block will be considered exhausted.