Stream Enrich: force jackson-databind to 2.9.3

Related to #3701 and #3702, we see a java.lang.NoSuchMethodError: com.fasterxml.jackson.databind.node.ArrayNode.<init>(Lcom/fasterxml/jackson/databind/node/JsonNodeFactory;Ljava/util/List;) error at runtime when using R103 (which is equivalent to master at the time of writing).

The solution for spark-enrich in #3702 fixes the problem in stream-enrich. That is, downgrading geoip2 to version 2.5.0 for stream-enrich. I am unsure if this is the correct approach. If it is, I can easily submit a PR.

Steps to reproduce:

git clone https://github.com/snowplow/snowplow
cd 3-enrich/stream-enrich
sbt "project kinesis" assembly
# Run target/scala-2.11/snowplow-stream-enrich-kinesis-0.16.0.jar with a config setup for Kinesis.
java -jar ...

We see ERROR com.amazonaws.services.kinesis.clientlibrary.lib.worker.Worker - Worker.run caught exception, sleeping for 1000 milli seconds! and the underlying error is java.lang.NoSuchMethodError: com.fasterxml.jackson.databind.node.ArrayNode.<init>(Lcom/fasterxml/jackson/databind/node/JsonNodeFactory;Ljava/util/List;).

[BenFradet]

I was able to find the root cause:

• our travis uses publishLocal for SCE and that includes jackson-databind-2.9.3
• when releasing we use publish and that includes jackson-databind-2.2.3 (c.f. http://search.maven.org/#artifactdetails|com.snowplowanalytics|snowplow-common-enrich_2.11|0.32.0|jar)

Apparently publishLocal relies on ivy2 whereas publish relies on maven and I guess they have different dependency resolution mechanisms.

[BenFradet]

Reopening til we get to the bottom of this.

Also you might be interested in https://github.com/snowplow/snowplow-docker 👍