Multiple data source problem in erroneous records

- After executing a mapping with multiple data sources, only the erroneous records from the main data source are saved. Records from secondary schemas are not saved.
- Also, reconsider using takeSample from spark in `writeErroneousDataset` method. Currently, we infer the schema by taking a sample from erroneous records assuming the schema of all records is the same. 