parse json before querying, output json column types

[huttotw]

This modifies the json_value macro to parse the JSON before attempting to extract a value. This causes the native JSON_QUERY function to return a column of JSON type instead of STRING which creates a better interface for users.

In short, now customers can query our event_properties, source_properties and user_properties using dot notation in BigQuery. Snowflake is unaffected.

SELECT
     user_properties.my_custom_property
FROM events;

instead of

SELECT
   JSON_QUERY(user_properties, '$.my_custom_property')
FROM events;

When the events table is built incrementally and this is persisted, it may enable some better performance by BigQuery since it's now aware the column is JSON (compression, etc).

[huttotw]

@camphillips22 the json_value function has changed with will ultimately affect the events model. @zerodelayy noticed that the column types in the event models were still be set as STRING in BigQuery. After examining the docs a little closer, JSON_QUERY only outputs JSON if the input is JSON, otherwise a STRING. Since BigQuery is synced with properties as a string, we needed to call PARSE_JSON prior to JSON_QUERY so that the return type of JSON_QUERY is JSON.

Before, you could still access JSON properties using the JSON functions provided by BigQuery, this is just a slightly more "correct" way to do it.