Add an error column containing decoding errors of each record

## Background
Spark data sources, like `csv`, create a column containing bogus records it was unable to parse.
Currently, `spark-cobol` just ignores all decoding errors and writes `null` to the target fields.

## Feature
It would be great if `spark-cobol` generated a column `decoring_errors`, containing all fields and values it failed to decode.
This should be optional (with the default being `false`) since generating this column can have an impact on performance and the output size.