Adds on_ending callback to allow processors to mutate spans before End operation

[pantuza]

This Pull Request aims to solve the issue #1695 by adding the on_ending callback to SpanProcessor class.
It also make sure to call that method during the span End operation (finish() method), right after the end
timestamp has been set but while the span is still mutable.

Based on the spec, this implementation complies with:

• The on_ending callback is optional within ANY SpanProcessor. Only called when exists.
• called during the span End() operation
• Span object is still mutable
• MUST be called synchronously within the Span End operation
• it should not block or throw an exception
• MUST guarantee that the span can no longer be modified by any other thread before invoking on_ending
• modifications are only allowed synchronously from within the invoked on_ending
• SpanProcessor on_ending callbacks are executed before any SpanProcessor's on_end callback is invoked

[dmathieu]

This is an experimental feature and may have breaking changes. How about specifying that here ?

[pantuza]

Thank you for the suggestion. I will update it.

[kaylareopelle]

Thanks for your PR, @pantuza! I want to think about this a little more, but I have some feedback on the documentation for now.

[pantuza]

Thanks for your PR, @pantuza! I want to think about this a little more, but I have some feedback on the documentation for now.

Thank you for your feedback and documentation suggestions. I really appreciate it.
Take the necessary time to think about the PR and please, let me know in case you think there is anything left to be done in this PR.

To give you a little background context, have started this work because of the following discussion in this Pull Request #1713

[robbkidd]

… have started this work because of the following discussion in this Pull Request #1713 [👀]

1713 is this pull request. Is there a different discussion you meant to link to?

[pantuza]

… have started this work because of the following discussion in this Pull Request #1713 [👀]

1713 is this pull request. Is there a different discussion you meant to link to?

You are right! Thank you for pointing it out. The correct discussion is in #1690

[mwear]

Overall, this looks good to me. For reference, Java already has this: open-telemetry/opentelemetry-java#6367. We chose to rename the end method to finish since end is a Ruby keyword. One thought I had was whether or not we should considering calling this on_finishing, but I am ok with this as is. Curious if anyone else has any thoughts.

[kaylareopelle]

Thank you!

[kaylareopelle]

One thought I had was whether or not we should considering calling this on_finishing, but I am ok with this as is. Curious if anyone else has any thoughts.

I do like the idea of using on_finishing here to keep the same term around, but don't feel strongly one way or the other. If we use on_finishing instead, we may want to add a link to the SDK documentation or "On Ending" to clear up any confusion that might exist connecting the on_finishing method to its specification.

[pantuza]

I also like the idea of using on_finishing to keep names concise with the project decisions.
If you don't mind I will update the code that way and make sure to document it properly.