fix(overlaybd): probe p2p before remote runtime initialization

[WindSongShaoGe]

Summary

• Probe the local P2P facade before initializing the shared remote runtime.
• Prevent the first remote layer reads from initializing the runtime without P2P acceleration.
• Use an unaccelerated registry path for P2P UUID probing to avoid recursively routing the probe through the P2P HTTP facade.
• Add regression coverage for remote runtime initialization order.

Problem

The remote runtime is stored in a OnceCell. Previously, P2P acceleration was only probed from create_image_file. If another call initialized the remote runtime first, the runtime permanently captured an empty acceleration address, so subsequent layer reads bypassed P2P acceleration.

Validation

• OverlayBD library tests: 310 passed, 0 failed, 1 ignored
• Clippy passed with warnings denied
• Rust formatting check passed
• git diff --check passed

Refs #138

[github-actions]

🔍 OpenCodeReview found 1 issue(s) in this PR.

• ✅ Successfully posted inline: 1 comment(s)

[github-actions]

enable_acceleration() performs synchronous DNS resolution and TcpStream::connect_timeout calls (up to one second per resolved address). Calling it directly while initializing this async OnceCell can block a Tokio worker whenever any remote-open path first initializes the runtime, delaying unrelated tasks. Run the probe via tokio::task::spawn_blocking (or replace it with an async, bounded probe) before setting the registry address.

Suggestion:

Suggested change

	self.enable_acceleration();
	let service = self.clone();
	tokio::task::spawn_blocking(move || service.enable_acceleration())
	.await
	.context("join acceleration probe task")?;