CDAP components should expose errors

Some CDAP plugins like Database do not expose error handling. When we process a pipeline in this case the errors may not always caused by database availability, for ex: size of the field not allowing data fit completely, errors on data encodings, PK/Unique keys violations, etc. In those cases it is makes sense to continue processing and simply handle errors via separated pipeline branch for further handling. I believe most components should expose this and let developer decide how to handle errors.