fix: detect early PTY exit in daemon sessions (#320)

[Wirasm]

Summary

When using kild open --daemon or kild open --resume --daemon, the daemon PTY session sometimes exits immediately after spawning. The kild open command reported success (session set to Active), but kild attach failed with the confusing error "session not running" — with no diagnostic info about why.

• Added a 200ms post-creation health check that polls the daemon after PTY spawn. If the PTY already exited, the error is caught immediately with exit code and scrollback output
• Added exit_code field to DaemonSession and SessionInfo so the exit code is stored and surfaced via IPC
• Added get_session_info() and read_scrollback() daemon client functions for the health check
• Applied the same fix to both create_session and open_session daemon paths

Root Cause

No post-creation health check existed in the daemon code path. create_pty_session() returned as soon as the daemon spawned the PTY, and the exit code was logged but discarded. Fast-failing processes (bad resume session, missing binary, env issues) would exit within milliseconds, but kild-core saved the session as Active and returned success.

Changes

File	Change
crates/kild-daemon/src/session/state.rs	Added exit_code field, getter, setter to DaemonSession
crates/kild-daemon/src/types.rs	Added exit_code to SessionInfo wire type
crates/kild-daemon/src/session/manager.rs	Store exit_code in handle_pty_exit before set_stopped
crates/kild-core/src/daemon/client.rs	Added get_session_info() and read_scrollback() functions
crates/kild-core/src/sessions/errors.rs	Added DaemonPtyExitedEarly error variant with diagnostics
crates/kild-core/src/sessions/handler.rs	Added 200ms health check in both daemon code paths

Testing

• cargo fmt --check — clean
• cargo clippy --all -- -D warnings — 0 warnings
• cargo test --all — all tests pass
• cargo build --all — clean build

Validation

cargo fmt --check && cargo clippy --all -- -D warnings && cargo test --all && cargo build --all

Issue

Fixes #320

[Wirasm]

Self Code Review

Summary

Clean implementation that directly addresses the root cause. No critical issues found.

Findings

Strengths

• Root cause correctly addressed: 200ms health check catches fast-failing PTY processes before user sees confusing "session not running" error
• Exit code plumbing is clean: field, getter/setter, wire type, storage in handle_pty_exit — all follow existing patterns
• DaemonPtyExitedEarly error provides excellent diagnostics (exit code + last 20 lines of scrollback)
• Backward-compatible wire protocol change (skip_serializing_if)
• Both create_session and open_session daemon paths covered

Edge Cases Handled

• Daemon unreachable during health check → proceeds without check (no regression)
• Empty scrollback → empty string in error message (exit code alone still useful)
• Base64 decode failure → empty vec via unwrap_or_default()
• Stopped daemon session → cleanup via destroy_daemon_session(force=true)

Security

No security concerns identified.

Checklist

• Fix addresses root cause from investigation
• Code follows codebase patterns
• Tests cover the change
• No obvious bugs introduced