fix: inconsistent interceptor span emission when workflow is in replay [WIP]

[sourcec0de]

What was changed

This PR addresses an issue that causes RunWorkflow:* spans to be emitted when workflows are viewed in the UI (i.e in replay). To make the behavior consistent, startNonReplaySpan was extracted from tracingWorkflowOutboundInterceptor and simplified to check the workflow context and return a no-op span when the workflow is in replay mode. Then startNonReplaySpan is reused in all places where a workflow context is in scope.

Why?

Discern uses the RunWorkflow:* otel span to alert on complete workflow failures (i.e a workflow that has died indefinitely). When viewing the temporal UI in the cloud console, our monitoring is re-triggered because the query to the worker replays the workflow history and re-emits the RunWorkflow span.

Historical context in slack

How was this tested:

I am still in the process of testing this using go workspaces by aliasing my branch of temporalio/sdk-go in our Go environment.

[CLAassistant]

All committers have signed the CLA.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[sourcec0de]

cc: @cretz