fix: make Core restart failures explicit

[Ezio1997-K]

Summary

Make Core startup/restart failures explicit and preserve useful diagnostics:

• propagate the 5-second IPC readiness timeout as a typed failure instead of logging and continuing;
• guard Core startup state with try/finally, so isStarting always resets;
• clean up a half-started Core while preserving the original startup error;
• make initial preload() wait for the first Core startup result;
• fail when Core initialization/state setup returns false;
• clear stale groups/providers at the start of a desktop profile hard switch;
• keep logs and requests instead of erasing the incident immediately after a switch;
• explicitly catch asynchronous profile-change failures;
• document that checkCoreHealth() checks only IPC/control-plane health, not TUN or mixed-port data-plane health.

This complements #209 but does not change profile apply ordering and does not add automatic restart/watchdog behavior.

Why

In #330 the App, Core process, Helper and mixed-port listener can remain alive while TUN/mixed traffic is unavailable. Treating socket connection or getIsInit as proof of full health leaves the UI showing stale nodes/status and hides restart failures.

The change deliberately avoids automatic self-healing restarts: a failed restart can extend an outage, and silently restarting would erase useful evidence.

Verification

• dart format on all five modified Dart files: clean;
• dart analyze lib/clash/core_startup.dart: no issues;
• standalone Dart 3.12.2 smoke coverage: normal completion, typed timeout, concurrent-start rejection, and state reset after success/failure: PASS;
• git diff --check: PASS.

Not run locally:

• full flutter test / whole-project analyze;
• Windows build or end-to-end TUN test.

The local environment did not have Flutter installed; this limitation is intentionally reported rather than treating the standalone Dart checks as a full Flutter build.

Related: #330, #209

[appshubcc]

部分已合并

其中过度设计的逻辑未添加