Define explicit LLM stream event seam

[kitlangton]

Summary

• Defines an opencode-owned LLM.Event union for the existing LLM.Service.stream seam.
• Keeps the current AI SDK runtime path intact, with a single cast at the adapter boundary.
• Adds a processor test that consumes explicit LLM.Event values directly.

Tests

• cd packages/opencode && bun typecheck
• cd packages/opencode && bun run test test/session/processor-effect.test.ts
• cd packages/opencode && bun run test test/session/message-v2.test.ts
• cd packages/opencode && bun run test test/session/retry.test.ts
• cd packages/opencode && bun run test test/session/prompt.test.ts
• cd packages/opencode && bun run test test/provider/transform.test.ts
• git diff --check

Notes

• No native @opencode-ai/llm runtime integration yet.
• No AI SDK dependency removal in this slice.
• test/session/compaction.test.ts is still baseline-failing on clean origin/dev with AsyncFiberError: An asynchronous Effect was executed with Effect.runSync.

[kitlangton]

Closing in favor of #26639, which tests the migration path using @opencode-ai/llm LLMEvent as the session stream seam instead of preserving an AI SDK-shaped wrapper type.