Unmodified field tokens are re-tokenized in UPDATE statements

[samsternberg]

snowflake_functions/snowflake_udf_demo/skyflow_udf_setup.sql

Line 443 in f0bce64

for record in stream_records:

Snowflake UPDATE statements are written to Snowflake Streams as a full DELETE and full INSERT. Since the INSERT contains all fields in the record, including those already tokenized, these token values are re-tokenized. This should not happen and only the updated fields (which contain new PII) should be tokenized.