Skipping null fields in Big Query inserts from Dataflow

[damccorm]

I am just wondering if we can skip the null fields from the JSON messages to reduce the serialization overhead and also making the big query insert faster by reducing the query size?

Imported from Jira BEAM-10070. Original Jira may contain additional context.
Reported by: varun.sharma.