LOAD DATASET on data with no schema produces an unhelpful error message.

[okennedy]

What pain point is this feature intended to address? Please describe.
When importing JSON data, if schema inference fails, an unhelpful error message appears.

Vizier tries to get the corrupted records field out of the JSON data, but if schema inference fails, it's the ONLY field being exported. Spark does not allow queries over ONLY the corrupted record field, so we get an error message about that

Describe the solution you'd like
It would be more helpful to get an error message to the effect of 'your JSON data is bogus, and I can't infer a schema... maybe try turning on multiline imports?'