This is the refreshed PR for #124, the old one was stale for too long and cannot be cleanly merged.
This PR does following things:

1. Add a --system-properties command line parameter to the Spark job, so the client can pass multiple secrets to the job.
2. The reading from JDBC function already exists in the job code but doesn't have public interface to use, now this function is enabled when you have JdbcSource in the feature definition file.
3. On the client side, a JdbcSource is added to represent a data source from JDBC, with auth, and corresponding data model change.
4. Various fixes to both job and client sides.
5. An E2E test to get offline features from JDBC data source.

Currently the E2E test cannot run in the CI env because it requires a database with green_tripdata_2020-04.csv imported into a table called green_tripdata_2020_04, with auth configured correctly.

To run the E2E test locally, you need:

• Create a SQL database instance with user/pass auth on Azure, AAD/token auth is not covered by this test.
• Import green_tripdata_2020-04.csv into the database with the table name set to green_tripdata_2020_04, be aware of the column type, all numeric columns like prices or distances must be "FLOAT" and all ID columns must be "INT" if you're using AzureSQL.
• Set following environment variables
  • nycTaxiBatchJdbcSource_USER: The user name to login to database server
  • nycTaxiBatchJdbcSource_PASSWORD: The password to login to database server

Then you should be able to run the test_sql_source.py directly or via pytest, both Databricks and Azure Synapse should work. You can use whatever user/pass combination for each Jdbc data source, not limited to the previously global unique JDBC_* settings.