feat(sandbox): keep mount endpoints stable

[VAIBHAVSING]

• Add a stable authenticated per-VM relay for native sandbox filesystem mounts.
• Share lazy single-flight lifecycle state between filesystem and process binding operations, including idle replacement and credential rotation.
• Add replacement coverage and document the updated sandbox lifecycle.

Closes #1826

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 92ab1b7d88

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Avoid awaiting in-flight sandbox startup during dispose

When dispose() is called while the first provider-backed sandbox operation is still inside provider.start(), this await makes VM teardown depend on that startup finishing. With the documented startupTimeoutMs: 0 setting, a hung provider start makes AgentOs.dispose() hang forever; even with the default timeout, the disposed-path startup promise later rejects after cleaning up the late client, so disposal reports a startup failure rather than completing cleanup.

Useful? React with 👍 / 👎.