Use HDFS file system based on some parameter (schema or `spark.defaultFS`)

Currently the HDFS namenode configuration is hardcoded to localhost:9000 which is not usable for the end users. Make possible for native code to read spark configuration