Adding missing script_location parameter while creating a Spark Command

[rohan-viz]

Closes #86

Although it is possible to submit a Spark command with a script location, this is not achievable through the SDK. This scenario is a very popular use case and these two are examples of important use case.

1. When the program or script is long enough forming a request is painful using REST.
2. A lot of times people already have scripts in s3 which are being used by other kinds of jobs. They always want to use the existing scripts and it becomes easy for them to just refer the path.

Rest API already supports this, refer here -
https://docs.qubole.com/en/latest/rest-api/command_api/submit-a-spark-command.html

[rohan-viz]

closes #100
If I hit qubole API to create schedule directly with an invalid value of frequency say -1, I get the following response body.

{
"error": {
"error_code": 422,
"error_message": "Validation failed: Frequency must be an integer greater than 0"
}
}
But while using the qubole sdk,we get the response as javax.ws.rs.client.ResponseProcessingException: javax.ws.rs.ClientErrorException: HTTP 422 Unprocessable Entity

Hence somewhere you have lost the exact error message given by qubole API .
422 status code can be given by qubole for multiple reasons.

So in ErrorResponseFilter class, we are reading this error response from the input stream. Hence the client using the SDK will not be able to see the exact error message. It needs to be stick back to the response context.