[FEATURE REQUEST]: Expose JVM exceptions

Currently an exception is only logged and then translated per the below.

[2020-04-03T00:31:33.1953860Z] [0ecc54d4f7e9] [Error] [JvmBridge] JVM method execution failed: Nonstatic method sql failed for class 435 when called with 1 arguments ([Index=1, Type=String, Value=select
 NOTACOLUMN from tmp], )
[2020-04-03T00:31:33.1955104Z] [0ecc54d4f7e9] [Error] [JvmBridge] org.apache.spark.sql.AnalysisException: cannot resolve '`NOTACOLUMN`' given input columns: [tmp.Parameter]; line 1 pos 7;
'Project ['NOTACOLUMN]
+- SubqueryAlias `tmp`
   +- Relation[Parameter#20315] parquet<...SNIP...>

Is it possible maybe throw an AggregateException with both exceptions or pass an InnerException containing the full original exception?

[thrixton]

Some more thoughts. #541 is quite a basic implementation and will expose the relevant information, but not in the nicest format. @imback82, would it be possible or advisable to expose more information from the JVM? e.g.

Field	Example
ExceptionType	org.apache.spark.sql.catalyst.parser.ParseException
Message	mismatched input ‘fu’ expecting {‘(’, <…SNIP…>'}(line 1, pos 0) == SQL == fu
StackTrace	org.apache.spark.sql.catalyst.par<…SNIP…>

Even better would be transposing most / all relevant exceptions from the JVM to c# exceptions.
To give some context, at some point I will have to parse the exception message so I can draw out specific error details, e.g. if the exception is an SQL exception (it is above), where it occurred (line 1, pos 0).
This task would be a lot easier with more exception details.

This would involve changes on the scala side though, I believe.

Thoughts?

[imback82]

Thanks @thrixton. Can you make the change and show how the new log message will look?