[Bench] Improvements for io_encode_benchmark.cc

[CaiusDai]

1. Current data generation can only control the cardinality and the data sampling scope is quite broad. We need more data generation function to test features for different encoding approach. (Like data range for Delta Bit Packing)
2. We need to configure the row group size to test how it would impact different encoding.
3. Consider change the data loading logic to make it the same for sort-read benchmark.
4. We need scripts that can run benchmark and draw graphs based on the results.