fix(capture): recover stale traffic takeover state

[MiChongs]

Summary

• persist an fsync-backed recovery journal and advisory lock before mutating host networking
• recover stale nftables/iptables rules, policy rules, route tables, and TUN state at the next startup
• distinguish Android VpnService mode from root-managed Android/Linux traffic takeover
• include SIGTERM in the ordered shutdown path

Root cause

Traffic cleanup depended on the process reaching the cooperative asynchronous stop path. A crash or forced termination could leave kernel networking rules installed, causing the host—especially a rooted Android device—to lose connectivity.

Impact

A successor process now self-heals stale takeover state before installing new rules, while the advisory lock prevents two instances from managing the same host state concurrently. SIGTERM performs immediate ordered cleanup; uncatchable termination such as SIGKILL is recovered on the next startup.

Checks

• cargo check -p wuther-core
• cargo test -p core-capture --lib platform::linux_recovery::tests
• full cargo test -p core-capture --lib: 313 passed during implementation

Linux cross-compilation reached third-party native dependencies but could not complete because the local environment lacks a Linux C sysroot (for example assert.h and sys/types.h).