Schema inference meta-ticket

[chuwy]

Just dumping here what we learned at Vienna hackathon.

Roadmap

1. Iglu Core: unversioned schema Scala Core: add support for Unversioned Schema #147
2. Iglu Core: split SchemaKey into data- and schema-specific Scala Core: split SchemaKey into data- and schema-specific #167
3. Schema DDL: version bump mechanism. Don't think this should be a part of Schema DDL, but this work relies on Schema AST included in this lib
4. Schema Guru: publish as a library
5. Scala Registry: add Schema Guru and its inference mechanism
6. Iglu Client: support for unversioned data

Outstanding questions

1. Where in the pipeline data becomes properly versioned
2. How should we decide if this is new model or non-sense data
3. Should we introduce some other Schema/DDL formats and introduce notion of origin before starting to work on it