Adds spark section to README.

[mansu]

Updated README explaining the steps to run the task as a local spark job.

[codefromthecrypt]

We should be careful about line and breakout count for a top-level readme. We already explain how to submit the job in two ways.

More importantly, I'm not sure this is equivalent. Particularly, the main class collects properties and derives the jars to submit. I'm not sure this will do the same. Can you test this?

After testing, whatever text should have the same results (probably same args) as the other two examples in the "Run the job" section. That way, users can map in their head what is different.

If this is works and is an alternative to the "java -jar" approach, we could consider putting this in the zipkin-sparkstreaming-job/README.md since it requires more parameters to do the same thing (since the main class has to be explicit in this case). If that was done, we'd add another link in the "Run the job" section to zipkin-sparkstreaming-job/README.md for detailed info. which prevents the README from becoming very long (very long readmes can throw off people)