fix(session): mark session Stopped even when daemon is unreachable

[Wirasm]

Summary

• Consolidate is_daemon_unreachable helper from list.rs into DaemonClientError::is_unreachable() method on the error type itself
• In stop_session(), classify daemon errors at collection point — unreachable errors (NotRunning, ConnectionFailed, ProtocolError, Io) are treated as "PTY already dead" with warn! instead of being pushed to daemon_errors
• The early return at line 124 now only fires for real DaemonError protocol errors where the daemon is alive
• Apply same fix to stop_teammate() which had identical unconditional error mapping

Test plan

• cargo fmt --check passes
• cargo clippy --all -- -D warnings passes
• cargo test -p kild-core passes (5 new tests for error classification)
• Manual: kild create test --daemon → kill daemon → kild stop test → session marked Stopped

[Wirasm]

PR Review Summary

Reviewed by: code-reviewer, pr-test-analyzer, silent-failure-hunter, type-design-analyzer, code-simplifier
Skipped: docs-impact-agent (internal refactor, no public API/config/docs changes)

Critical Issues (0 found)

None.

Important Issues (0 found)

None.

Suggestions (1 found)

Agent	Suggestion	Location
code-reviewer	destroy.rs:219 only special-cases NotRunning for daemon errors, not all unreachable variants (ConnectionFailed, ProtocolError, Io). Consider adopting is_unreachable() there too for consistency. Not blocking — separate concern, separate PR.	crates/kild-core/src/sessions/destroy.rs:219

Strengths

• Clean consolidation — moving is_daemon_unreachable from a private function in list.rs to DaemonClientError::is_unreachable() follows SRP and makes the classification reusable across list.rs, stop.rs, and any future consumer
• Safe default for future variants — the !matches!(DaemonError { .. }) approach means new error variants are classified as unreachable by default, which is the correct safety posture
• Tests fully migrated — all 5 classification tests moved to their natural home on the error type with no coverage loss
• stop_teammate fix — previously any daemon error was a hard failure; now unreachable errors are gracefully handled, matching the intent that "daemon dead = PTY dead = stop succeeded"
• Event naming — new _unreachable events follow the established {layer}.{domain}.{action}_{state} convention

Verdict

READY TO MERGE

Clean, focused refactor that fixes a real bug (stop failing when daemon is unreachable) with proper error classification, good test coverage, and no regressions.