Document the status of the attribute registry and whether md or yaml are the source of truth

[lmolkova]

Based on the discussion in the #1345 it seems we don't have clarity on:

• what's the source of truth yaml or md
• what represents the registry - https://github.com/open-telemetry/semantic-conventions/blob/main/docs/attributes-registry or https://github.com/open-telemetry/semantic-conventions/tree/main/model/registry
• how attribute definition is decoupled from usage (by folder structure or dedicated schema Semconv schema: formalize registry definition #1269)

This becomes much more interesting when we combine it with code-generation and semconv artifact structure.

I believe we need to think through and document answers to these questions

[lmolkova]

I propose treat resolved conventions as the source of truth - resolved conventions are used for MD and code generation, the original yaml files organization/schema can change without affecting any consumers.

[joaopgrassi]

I propose treat resolved conventions as the source of truth - resolved conventions are used for MD and code generation

+1. And by "resolved" here we means the ones in inside model/ (excluding the registry).

what represents the registry - main/docs/attributes-registry or main/model/registry

The Registry is just the registry. All that's available to be used and defined. To me they are not semantic conventions, because they are not put into any context. Only once they are put into usage/context, they become semantic conventions to me. Maybe that's a direction we want to go to better define the role of the registry.