Support for pluggable backend storage

[tookko]

Currently Karapace requires a Kafka cluster for storing schemas. While the topic name can be changed, it creates management / mapping overhead in cases where multiple schema registries are needed, or when schemas are needed outside Kafka, or when dealing with limited HW resources.

Suggestion is to abstract backend storage details and allow different storage plugins to be developed (e.g. relational DB or k/v store).

[jjaakola-aiven]

The https://github.com/Aiven-Open/karapace/blob/main/karapace/in_memory_database.py has an API that can be extended for pluggable backend.

The API can be extracted to abstract base class. The point where some extra care is required is the get_schema_id function, when relational database can provide monotonically increasing identifier. Also some locking that protects the schema id generation can also be removed when using database as storage.