Properly handle empty csv values when converting to json

[jeffadair]

Describe the bug
When running in csv-to-* (originally seen with csv-to-sqs), the records get added to the with the empty csv fields as "NaN". This is not proper json and causes json parse errors in stream-loader

To Reproduce
Steps to reproduce the behavior:

1. Use a docker-compose-demo demo yaml.
2. Edit the producer definition to use csv-to-sqs (rabbitMQ and Kafka should also see the problem)
3. Use the following input file for the producer: SENZING_INPUT_URL: /opt/senzing/g2/python/demo/truth/truthset-person-v1-set1-data.csv
4. Add the following to the producer service in the yaml file:
   volumes:${SENZING_G2_DIR:-/opt/senzing/g2}:/opt/senzing/g2
5. Start the formation as normal, watch the output from stream-loader. You should see errors like this:
   senzing-stream-loader | 2021-02-04 23:30:55,655 senzing-50010889E G2Engine.addRecord() G2ModuleGenericException: 30121E|JSON Parsing Failure [code=3,offset=114]; JSON: {"AMOUNT": 200, "CELL_PHONE_NUMBER": "982-122-2332", "DATA_SOURCE": "TEST", "DATE": "1/24/2018", "DATE_OF_BIRTH": NaN, "DRIVERS_LICENSE_NUMBER": NaN, "DRIVERS_LICENSE_STATE": NaN, "EMAIL_ADDRESS": "mickey@mmail.com", "ENTITY_TYPE": "GENERIC", "GENDER": NaN, "HOME_ADDR_CITY": "Andersonville", "HOME_ADDR_COUNTRY": NaN, "HOME_ADDR_LINE1": "P.O. Box 12987", "HOME_ADDR_POSTAL_CODE": "60611", "HOME_ADDR_STATE": "IL", "HOME_PHONE_NUMBER": NaN, "NATIONAL_ID_COUNTRY": NaN, "NATIONAL_ID_NUMBER": NaN, "PASSPORT_COUNTRY": NaN, "PASSPORT_NUMBER": NaN, "PRIMARY_NAME_FIRST": "Marie", "PRIMARY_NAME_LAST": "Sanchez", "PRIMARY_NAME_MIDDLE": NaN, "RECORD_ID": 1097, "SSN_LAST4": NaN, "SSN_NUMBER": NaN, "STATUS": "Active"}

Expected behavior
Load without errors.