Ensure we have good coverage of csv and json encoders and decoders.

[nathanielc]

Once we have confidence in our coverage of the encoders through the units tests remove the unnecessary use of the encoder/decoders in the end to end tests.

[aanthony1243]

ideas:

1. Wrap the CSV encoder to take a block of text to diff, then we can possibly run the existing CSV tests to cover the encoder. These could run similarly to how we have it set up now: input.csv -> flux query -> output.csv
2. Wrap the CSV encoder + decoder together to ensure that production/consumption between the two are precise.