fix: ensure proper process termination when using --exit-no-conn or --once

[saltbo]

Problem

Resolve #1461

When using --exit-no-conn or --once options, ttyd would exit immediately without properly waiting for the child process to terminate. This could result in:

• Child processes not receiving the configured signal
• Incomplete cleanup of child process resources
• Orphaned processes in some cases

Root Cause

On certain systems (particularly macOS), the waitpid() system call may fail to properly reap zombie processes in some edge cases, causing the process_exit_cb callback to never be invoked, leading ttyd to wait indefinitely.

Solution

Implemented a dual-mechanism approach for robust process termination handling:

1. Primary mechanism: Preserve existing waitpid() and process_exit_cb logic
2. Fallback mechanism: Add uv_timer_t that periodically checks process status using kill(pid, 0)

If the primary mechanism fails, the fallback timer detects when the process no longer exists and triggers ttyd exit.

Changes

• Added exit_timer field to pty_ctx_t structure
• Implemented exit_timer_cb() for periodic process status checking
• Modified process cleanup logic to handle both normal and fallback exit paths
• Enhanced --once and --exit-no-conn behavior to properly wait for process termination

Testing

• Verified on macOS with various command types including asciinema
• Confirmed that configured signals are properly sent to child processes
• Tested that ttyd exits cleanly after child process termination
• No impact on normal operation when not using --once or --exit-no-conn

Compatibility

• Backward compatible, no breaking changes
• Only affects behavior when using --once or --exit-no-conn options
• Minimal performance overhead (timer only active during exit sequence)

Fixes the issue where ttyd would not properly wait for child process termination, ensuring reliable signal delivery and process cleanup.