Move to io.opentelemetry.semconv package

[jack-berg]

Related to #5, #2.

• SemanticAttributes now lives in a package which doesn't include *.trace.*, which is accurate given that includes attributes used across signals.
• ResourceAttributes lives in the same package as SemanticAttributes because the additional level to the package hierarchy serves no value.
• The package includes v1 to give allow us to make breaking changes to the code generation logic as discussed in comment. We are still publishing an -alpha artifact, so we are allowed to make breaking changes to the code generation logic, but I figure its best to keep the package change churn to a minimum, so trying to get us close to the structure we are likely to use when we publish a stable artifact. Discussed in a comment and decided to optimistically omit v1 from the package name. If later we need to make breaking changes after we release a stable version, we can skip to include v2 in the package name.
• The downside of this set of changes is that the package change will be a breaking change vs. the existing io.opentelemetry:opentelemetry-semconv artifact. I think that's fine tho and we can publish deprecation warnings that point to the new classes in io.opentelemetry:opentelemetry-semconv before deleting.

[trask]

• The package includes v1 to give allow us to make breaking changes to the code generation logic as discussed in comment.

what do you think of being optimistic and not including the v1 in the package name, and only adding v2 in the future if needed?

[jack-berg]

The package includes v1 to give allow us to make breaking changes to the code generation logic as discussed in #2 (comment).

what do you think of being optimistic and not including the v1 in the package name, and only adding v2 in the future if needed?

I'm fine with that too. Was hoping this PR would spur that conversation 🙂