Fix problem that caused CI on forks to fail

[jklukas]

We were trying to build an uberjar and stage in S3,
but that requires encrypted env vars that aren't avaialable to
builds on forks, so we need to make the assembly and deploy stages conditional.

[fbertsch]

LGTM after successful branch build

[codecov-io]

Codecov Report

Merging #436 into master will decrease coverage by 51.58%.
The diff coverage is n/a.

@@             Coverage Diff             @@
##           master     #436       +/-   ##
===========================================
- Coverage   69.62%   18.03%   -51.59%     
===========================================
  Files          48       48               
  Lines        4141     4141               
  Branches      146      146               
===========================================
- Hits         2883      747     -2136     
- Misses       1258     3394     +2136

Impacted Files	Coverage Δ
...ozilla/telemetry/utils/ConsistentPartitioner.scala	0% <0%> (-100%)	⬇️
...n/scala/com/mozilla/telemetry/ml/AMODatabase.scala	0% <0%> (-100%)	⬇️
...m/mozilla/telemetry/utils/CustomPartitioners.scala	0% <0%> (-100%)	⬇️
...cala/com/mozilla/telemetry/utils/Aggregation.scala	0% <0%> (-92.31%)	⬇️
...n/scala/com/mozilla/telemetry/utils/MainPing.scala	5.18% <0%> (-90.16%)	⬇️
.../main/scala/com/mozilla/telemetry/utils/udfs.scala	0% <0%> (-89.19%)	⬇️
.../mozilla/telemetry/views/GenericLongitudinal.scala	0% <0%> (-86.85%)	⬇️
...m/mozilla/telemetry/experiments/Permutations.scala	0% <0%> (-85.72%)	⬇️
...m/mozilla/telemetry/utils/SyncPingConversion.scala	0% <0%> (-81.82%)	⬇️
...n/scala/com/mozilla/telemetry/avro/JSON2Avro.scala	0% <0%> (-80.96%)	⬇️
... and 24 more

---

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 d41b87f...985f42f. Read the comment docs.

[jklukas]

https://travis-ci.org/mozilla/telemetry-batch-view/builds/389414524 was successful, so merging. The codecov report is erroneous, I believe, because one of the jobs died and had to be restarted, which confused the merging of reports.