Fix spark mode on experiment jobs

[sunahsuh]

Welp. I'm gonna guess at least a few of the problems with the experiment jobs are due to hard-coding local spark mode.

[codecov-io]

Codecov Report

Merging #274 into master will decrease coverage by <.01%.
The diff coverage is 50%.

@@            Coverage Diff             @@
##           master     #274      +/-   ##
==========================================
- Coverage   66.12%   66.11%   -0.01%     
==========================================
  Files          37       37              
  Lines        3371     3373       +2     
  Branches       80       83       +3     
==========================================
+ Hits         2229     2230       +1     
- Misses       1142     1143       +1

Impacted Files	Coverage Δ
...zilla/telemetry/views/ExperimentAnalysisView.scala	44.44% <0%> (+1.89%)	⬆️
...ozilla/telemetry/views/ExperimentSummaryView.scala	19.23% <0%> (-7.19%)	⬇️
...in/scala/com/mozilla/telemetry/utils/package.scala	81.03% <100%> (+1.78%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b4ecef0...55f7a3c. Read the comment docs.

[acmiyaguchi]

This makes sense, now I understand why the job for a single experiment finished in the same amount of time on 1 machine and 10 machines.

[acmiyaguchi]

I'm going to rerun the experiments job for 20170724 on airflow.