Revisit API error cases

[wwerner]

We're still experiencing intermittend hanging requests.
To approach this in a structured kind of way, I suggest writing a dedicated negative test suite for

• Trying to access non-existing resources (see also Accessing non-existing resources just hangs #96) -> 404
  • Organization
  • Unit
  • Context
  • Schema
  • Schema Version (by /organization and /version/<reference>)
  • Sources
• Syntactically invalid specifications (by POST, PATCH, PUT) -> 400
• Lifecycle state changes to invalid states -> 400
• Incomplete entities (by POST, PATCH, PUT) -> 400
  • Organization
  • Unit
  • Context
  • Schema
  • Schema Version (by /organization and /version/<reference>)
• Invalid schema version updates
  • Conflicting previous and current versions (gaps, same version, version no decreases) -> 409
  • Incompatible specifciations in compatible version update (for patch and minor updates) -> 409

[VaughnVernon]

@wwerner A lot of this can be avoided by projecting a CQRS query model to eliminate navigational queries. Instead we can simple take all the path parameters and make a single composite key that resolves to the schema version (or whatever level the user wants).

Key: org:unit:context:schema:version
Value: schema-version-definition

We can create projections even though we are currently not event sourced. All of the events are already there, or you can use full-entity state-based projections.