Update PK sequence value to 2**31 in clean-up step, not swap #49
Description
In #47, special handling for the edge case of a bigint conversion where the PK sequence has been set to negative (presumably after positive PK values were exhausted) was added. In that case, the sequence is reset to the first positive bigint value not within the integer range, which is 2**31; this happens during the swap step.
However, switching to PK values that exceed the integer range during the swap step means that the trigger that keeps the old table (which has an integer PK) updated will fail.
To address this, the sequence value needs to be updated from negative to 2**31 during the cleanup step, after the old table and update trigger function have been dropped.