Clarify logging around and usage of read_size startup option

The following error message:
"Cannot process row that is bigger than the IO size (row_size=X MB, null_indicators_size=0). To run this query, increase the IO size (--read_size option)."
https://github.com/cloudera/Impala/blob/cba02a9a8b8cb765f9f4d2884f0a45e9af1bf2c7/common/thrift/generate_error_codes.py#L209-L211

, and the read_size option in general, needs to be reviewed to reflect changes from IMPALA-3200.