fix(io): catch ValueError in _SafeWriter for closed file handles

[Bartok9]

Summary

Fixes #2428 — _SafeWriter.isatty() raises ValueError on closed file during subagent ThreadPoolExecutor teardown.

Problem

When subagents run in ThreadPoolExecutor threads, the shared stdout handle can close between thread completion and KawaiiSpinner.stop() cleanup. Python raises ValueError (not OSError) for I/O operations on closed files:

ValueError: I/O operation on closed file

The _SafeWriter class (added in a8409a1) only catches OSError, missing this case.

Root Cause

Two commits that interact:

1. a8409a1 (Mar 11) — Added _SafeWriter to catch OSError for broken pipes
2. 6e2be33 (Mar 20) — Added isatty() check in KawaiiSpinner.stop()

The isatty() call during spinner cleanup can fail with ValueError when the underlying handle is already closed.

Fix

Add ValueError to exception handling in write(), flush(), and isatty() methods:

except (OSError, ValueError):

This is a standard Python pattern — ValueError: I/O operation on closed file is the expected exception for operations on closed file descriptors, distinct from OSError which covers broken pipes and device errors.

Testing

• Verified the code paths handle both exception types
• This is a straightforward exception expansion with no behavioral change for healthy streams

[teknium1]

Merged via PR #2466 — cherry-picked with authorship preserved. Thanks!