Major Description
pandas and pandas-on-Spark both have an argument named `mode` in the DataFrame.to_csv, but the acceptable strings are different.
pandas can accept the "w", "w+", "a", "a+" where as pandas-on-Spark can accept "append", "overwrite", "ignore", "error" or "errorifexists".
We should map these acceptable strings to pandas.
e.g. "w" will work as Spark's "overwrite". In addition, mode can take Spark's "overwrite" too.
