Split lzo serializers into a separate sbt project

[istreeter]

This is a breaking change, which means our standard s3-loader assets (the jar file and docker image) will not be able to output lzo compressed files.

Instead, users will need to pull the -distroless tagged docker images, e.g.

docker pull snowplow/snowplow-s3-loader:2.2.0-lzo

We are doing this because the lzo library depends on hadoop version 2, which is now very old and not so actively maintained. If high vulnerabilities are reported in hadoop version 2, then we have no easy way to mitigate them in s3 loader. But by confining lzo compression to a separate image, we can more easily maintain a "clean" image, free from security vulnerabilities, and OK to output gzipped files.