Hi!

I would like to add the possibility to not register a new schema when encoding a message and passing the schema_name option only.

The motivation for this is that most of the times the schema evolution (registering and updates) are done during the Continuous Delivery process (during deployments, right before putting the new application code live), since this process is very delicate and we do not want to risk accidentally evolving the schema when trying things out in the development phase, I think it's important to have the option to tell the library to use a register_schemas false mode while encoding.

In my company, currently the applications do not have to keep track of the current schema_id or even the version of the schemas used. So generally we call the encode method with the following options:

avro.encode(message_payload, schema_name: the_schema_name, validate: true)

The problem with this is that, if by any chance we're testing things out locally, there is a risk to trigger an evolution of the schema by registering a new version that was not ready to be evolved.

In order to avoid that, we want to provide the register_schemas option, which only in case of false checks if the schema is already registered (by using the existing check action) and caches the schema so that it's only checked once, if it exists, it returns the schema_id from there instead of attempting to register it.

This change is 100% backwards compatible, as the current behavior only changes if the user decides to pass the register_schemas option set to false.