Spanprocessors / TracerFactory

[z1c0]

Update SDK spec regarding span processors and Tracers / TracerFactories.

[SergeyKanzhelev]

What's the motivation of this limitation? I understand that it simplifies the implementation, but from the scenarios perspective I can imagine scenarios when user wants to enrich data or apply a custom filter for a specific tracer. Is it not a valid scenario?

Also, is the same limitation applied for samplers?

[z1c0]

This is actually not about simplifying the implementation but to have a well-defined and comprehensible specification how all Tracers created by the same Tracer factory must behave. It's less confusing if Tracers created by the same factory also share the same properties and configuration, hence behavior.
In the case where one might want to treat a Tracer differently:

• An additional TracerFactory can be used.
• SpanProcessors can use the tracer name and version for that.

I will look into Sampling and Propagators separately to have more focused PRs.

[SergeyKanzhelev]

If we will find out that per-Tracer processors are important we can introduce them later.

[SergeyKanzhelev]

This PR has enough sign offs from diverse set of companies. And it was active for at least 2 business days. Merging