fix(stream): persist mid-stream failures

[rdself]

Summary

This patch fixes a second stream lifecycle gap on top of the existing v3.8.26 stream readiness work: mid-stream transport/transform failures could clear the active request from memory without going through the chat completion finalization path. In that case the Logs page entry disappeared instead of becoming a durable failed call log, even though the upstream request may already have run for a long time.

Changes

• Route pipeWithDisconnect stream errors through a new controller onError callback so chatCore can own request finalization when appropriate.
• Persist non-200 stream completions with status, error, error code, payload summary, and usage metadata instead of clearing pending requests directly.
• Keep in-memory completed log detail status/error fields aligned with persisted call logs.
• Let stream-level failure handlers signal that they already handled pending cleanup, avoiding double-finalization.
• Merge Claude/Anthropic usage incrementally so message_start input/cache tokens are not lost when later message_delta chunks only contain output usage.
• Preserve input_tokens / output_tokens in usage normalization for Anthropic-style streams.
• Add regression coverage for owned stream error cleanup and Claude usage merging.

The upstream request shape is unchanged, including Claude Code compatible provider payloads.

Validation

• node --import tsx/esm --test tests/unit/stream-handler.test.ts tests/unit/stream-utils.test.ts tests/unit/active-request-stream-chunks-lifecycle.test.ts
• npm run typecheck:core
• Local production run against the v3.8.26 branch with verbose stream logging enabled.

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[diegosouzapw]

Obrigado, @rdself! 🙏 Excelente follow-up do #3879 — fechar o segundo gap do ciclo de vida do stream (falhas mid-stream agora viram call-log de falha durável em vez de sumir da página de Logs) é exatamente o tipo de robustez que faltava. Apreciei muito que a extração do streamFailureFinalization.ts está de fato plugada (importada e usada pelo chatCore), e o merge incremental de usage Claude evita perder os tokens do message_start. Validei localmente: typecheck:core limpo, 75/75 testes de stream/usage, eslint 0 erros, file-size + any-budget OK. Entra no próximo release. 🚀