Read support for BINARY in ORC

ORC has its own BINARY type, which has some differences compared to STRING/VARCHAR/CHAR, as BINARY:

• doesn't use dictionary encoding
• doesn't store min/max values in the statistics

The c++ library uses the same ColumnReader as for STRING, so the implementation efforts should be minimal: https://github.com/apache/orc/blob/a9ec6a2e39ed71ef8a2d874df14700956aa847be/c%2B%2B/src/ColumnReader.cc#L1752