Use last_pending_finalizer_policy_start_timestamp instead of last_pending_finalizer_policy_start_num

[arhag]

Instead of last_pending_finalizer_policy_start_num we should use last_pending_finalizer_policy_start_timestamp.

Then if last_pending_finalizer_policy_start_timestamp <= latest_qc_claim_timestamp and active_finalizer_policy_generation != last_pending_finalizer_policy_generation for a block (call it block A) that has a strong QC on it, we know that this strong QC, if claimed in a block (call it block B), would prove that the last pending finalizer policy referenced by the last_pending_finalizer_policy_generation in block A would become the active finalizer policy in block B.

With this change latest_qc_claim_block_num might not even be necessary in level_3_commitments_t, but it may be best to not get rid of it yet.

There is no harm in using the timestamp version rather than the number version and it is more consistent with other places in our implementation where we prefer timestamp over block numbers.

Furthermore, committing to the last_pending_finalizer_policy_start_timestamp in the finalizer digest may be useful in proving finality violations in certain cases; however it may not be sufficient (we might require a last_pending_finalizer_policy_finality_digest as well).