Wide index records

[markpapadakis]

There is partial support for wide index records already, and in practice, it won't be needed unless you have really HUGE datasets, but we should implement it anyway. This should be trivial, and will come down to generating a new index from the old, where the new entries are {seqNum:u64, fileOffset:u32} instead of {relative SeqNum:u32, fileOffset:u32}.