fix(blaze): accept nsfs mount roots

[WeissonHan]

Component

blaze

What happened?

When a persistent Linux network namespace is bind-mounted under /run/netns, /proc/self/mountinfo contains an nsfs entry whose root is a namespace identifier such as net:[4026539640] rather than an absolute path.

Blaze currently rejects that valid entry while validating template storage boundaries. As a result, blazed cannot start or restart on a host that already has a named network namespace, including a restart needed to resume a hibernated sandbox.

Expected behavior: accept the opaque one-component root used by nsfs, while continuing to require absolute roots for ordinary filesystems and absolute mount points for every filesystem.

How can we reproduce it?

1. Create a named network namespace so Linux bind-mounts an nsfs namespace under /run/netns.

2. Start blazed with template catalog validation enabled.

3. Observe startup fail while parsing an entry with this shape:

   1062 592 0:4 net:[4026539640] /run/netns/example rw shared:681 - nsfs nsfs rw

Diagnostic output

mountinfo root and mount point must be absolute

Anything else?

The correction should remain inside the mount table parser. It does not change a public API or configuration contract.

[github-actions]

🔀 Issue Router

• Component: blaze
• Confidence: 99%
• Owners notified: @kongche-jbw
• No assignee was added; assignment records active implementation ownership.

问题明确涉及 Blaze 的 mountinfo 解析。

Evidence: 标题和正文直接指向 Blaze，并描述 blazed 的 nsfs 挂载表解析问题。

This routing was applied by the ANOLISA Issue Router through GitHub Actions.