fix(reporter): wait and retry when task output forwarding hits EAGAIN

[naokihaba]

Context

• relates: vp run: task output forwarding aborts with EAGAIN when stdout turns non-blocking mid-run (follow-up to #2165) vite-plus#2824
• resolves: vp run: output forwarding fails with EAGAIN when an inherited Node child sets stdout non-blocking #782

Reproduced on Vite+ v1.0.0 via https://github.com/naokihaba/vp-eagain-repro .

This happens when running two tasks concurrently while vp run outputs to a pipe, such as in CI logging.

Task A ──[Pipe ①]──→ vp ──→ vp's fd 1 ──[Pipe ②]──→ CI log collection
                                  ↑
Task B (Node) ────────────────────┘ Sharing the same output destination

1. Task B (Node) sets the shared output to O_NONBLOCK, also making vp's fd 1 non-blocking.
2. Task A outputs heavy logs, which vp writes to Pipe ②.
3. Pipe ②'s buffer fills up instantly.
4. vp's write returns EAGAIN instead of waiting.
5. vp treats this as fatal and kills all tasks.

This fix only affects cases where write or flush returns WouldBlock. Outputs to terminals, files, or blocking pipes never trigger WouldBlock, keeping the previous behavior unchanged. Common errors like reader disconnections are still returned as before.