[exec-server] Require filesystem helper readiness

[erichoracek]

Stack

• depends on [exec-server] Kill dropped filesystem helpers #25116

Summary

• emit and flush a readiness marker from the sandbox filesystem helper before it waits for request input
• require the parent process to receive that marker within 10 seconds before sending the filesystem request
• preserve the existing unbounded wait for the filesystem operation itself
• add focused regression coverage for marker consumption and startup timeout

Why

A sandbox filesystem helper can stall during process startup before reading stdin. Waiting for a flushed readiness marker gives the parent a precise startup-health signal: if the helper never reaches its Rust entrypoint, the parent returns an error and the dropped child is cleaned up by the kill_on_drop(true) behavior from #25116.

This intentionally does not apply an arbitrary timeout to filesystem work after startup.

Validation

• just test -p codex-exec-server (188 tests passed; benchmark smoke passed)
• just fmt
• just fix -p codex-exec-server
• git diff --check