fix: make telemetry more resilient to missing data

[cdoern]

What does this PR do?

it seems sometimes row is missing some data like root_span_id, omit these entries and log a warning

example failure: https://github.com/meta-llama/llama-stack/actions/runs/16368394907/job/46250869773

telemetry tests should now pass

[ashwinb]

cc @ehhuang

[leseb]

@cdoern can you share a failure example?

[cdoern]

@leseb : https://github.com/meta-llama/llama-stack/actions/runs/16368394907/job/46250869773 on the main branch, I can link this in the PR

[leseb]

Rather than working around the error, can we investigate why root_span_id is missing in the first place? Thanks

[bbrowning]

Yes we should figure out why it's failing, as this should be deterministic. The only valid case where we'd be missing a root_span_id is when we're part of a distributed trace started outside of Llama Stack server with the client specifying the right trace propagation headers.

With that span, root_span_id is NOT strictly required and validation requiring it is probably incorrect. See #2494 where some of this was fixed before, but I would expect this to either pass or fail consistently vs flake in this way.

[ehhuang]

Rather than working around the error, can we investigate why root_span_id is missing in the first place? Thanks

+1 I can disable the test again in the meantime

[ehhuang]

#2814