Is your feature request related to a problem or challenge? Please describe what you are trying to do.

Whilst working on #171 I found the logic in parquet::arrow::RecordReader very difficult to follow. This mainly stems from the fact it has a state where it is part way through "consuming" a record, and state is mutated in multiple places.

Describe the solution you'd like

It is possible to simplify the logic, at the cost of potentially scanning repetition levels multiple times. In practice this is already done as consequence of calling RecordReader::reset. This combined with the overheads of splitting the buffers makes me suspect the performance difference will be negligible.

Describe alternatives you've considered

The existing logic is not wrong, and so could remain unchanged.