fix(span-buffer): Copy the event_id field into attributes

[gggritso]

Closes STREAM-294. The event_id field should be copied into the sentry.event_id attribute, to copy what Snuba does. I'm not sure why this field is missing, is there a specific reason? I took the liberty of adding a ticket and putting up the PR 🙏🏻

[untitaker]

event_id will stop existing once transactions are gone. so looking forward we cannot support it.

[gggritso]

@untitaker I'm confused. People will continue sending transactions via old SDKs, those transactions will keep having event IDs, Relay will continue setting that attribute. I'm not sure how many people have alerts/dashboards/workflow that rely on it, but dropping event_id from the buffer literally causes data loss, no?

[untitaker]

I mean, I don't know how you use this attribute in the UI. But shouldn't the product end up working the same no matter how a user sends the data? if the product relies on event_id today to identify transactions, it will start to change behavior once the user upgrades the SDK. so some kind of migration has to be done.

[gggritso]

@untitaker okay, I talked with some other folks on the performance team, here's the scoop:

1. The Performance product uses the event_id for a bunch of important functionality. The biggest one is that in the Trace Waterfall, event_id is used to fetch a bunch of extra info (breadcrumbs, performance issues, etc.) from nodestore, from occurrences, and other places
2. There is a clear path forward to stop relying on event_id, but it's long-term. We can stop using it once performance issue occurrences, breadcrumbs and other telemetry makes its way into EAP. Until that happens, we need event_id to look those things up

So, I think the upshot is that we have to keep event_id handling in the buffer for now, because otherwise switching to the buffer will break a bunch of existing functionality. FWIW Shruthi confirmed that there's a clear path forward for migrating user queries and UI functionality off event_id in the future