refactor Spark fetcher

[rayortigas]

While preparing to try improving how Dr. E collects Spark data, I went through some of the existing Spark-related code and tried to clean it up:

• adding unit tests
• re-organizing code
• pruning code
• closing open connections

Let me know if you have any questions...

[akshayrai]

@rayortigas, could you take care of the Travis build checks? We want Dr. Elephant to work with java 6 also.

[rayortigas]

Thanks for the note @akshayrai. I didn't get any messages about the CI failures; can you enable notifications if they aren't already enabled?

Just curious, why do we want to preserve compatibility with Java 6?

[rayortigas]

OK, it looks like CI passed this time. Please let me know if you have any other questions.

[akshayrai]

@rayortigas, thanks for fixing it. We have been preserving java 6 just to ensure that people who are still running java 6 on their clusters can use Dr. Elephant.

[akshayrai]

+1, LGTM.

@rayortigas, before we merge, can you fix the small conflict.

[shankar37]

+1

[rayortigas]

Closing, now that #162 replaces the old Spark fetcher entirely.