Improve error handling for when two instances of hlink try to point to the same database

[riley-harper]

When two instances of hlink try to load data from the same database, a huge error is printed by Spark. One of the important lines printed looks like

ERROR XSDB6: Another instance of Derby may have already booted the database <database_path>

With Spark 2, I was able to catch the Python exception that's thrown and print a more useful error message, but the huge JVM stack trace was still printed. In Spark 3, I think that we can set the spark.sql.pyspark.jvmStacktrace.enabled configuration option to prevent Spark from printing these stack traces. This may have an effect on other places where we get large JVM stack traces too.