feat(daemon): emit chat spans via weave SDK LLM

[rgao-coreweave]

Summary

Migrate chat spans from emitChatSpan / emitChatSpansFromAssistantCalls to the SDK LLM via parent.startLLM(...) + llm.record(...), with post-hoc startTime/endTime so replayed spans match transcript timestamps.

Folds in the migration cleanup (per the stack plan, no separate cleanup PR): drops the now-unused genaiSpans span helpers, leaving genaiSpans.ts as a thin ATTR + formatting layer; removes the transitional migration-snapshot test and the helper-level unit tests whose coverage now lives in the SDK genai tests + genai-span-usage-tokens.

Companion SDK PRs: wandb/weave#7078 (post-hoc timing, merged), wandb/weave#7195 (setAttributes/addEvent, merged), wandb/weave#7077 (SubAgent.startLLM).

Test plan

• npm run build
• npm test (44 tests)

[rgao-coreweave]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• refactor(genaiSpans): drop migrated helpers; keep formatting only #79
• feat(daemon): emit chat spans via weave SDK LLM #78 👈 (View in Graphite)
• feat(daemon): emit subagent invoke_agent spans via weave SDK SubAgent #77
• feat(daemon): emit tool spans via weave SDK Tool #76
• feat(daemon): emit turn spans via weave SDK Turn #75
• feat(daemon): wrap claude code sessions in weave SDK Session #74
• feat(daemon): adopt Weave Node SDK for tracer provisioning #73
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.