Fix Java OOM by registering kill as TeardownFunc

[andy1li]

Note

Medium Risk
Changes process lifecycle and SIGKILL teardown for all shell-based tests; behavior is intentional but affects every stage using ShellExecutable on Linux.

Overview
Shell test harnesses now tear down PTY child processes reliably by wiring cleanup that was previously left as a TODO.

NewShellExecutable registers Kill() on stage teardown so the spawned program (and its children) are stopped when a test stage finishes. Kill() is a no-op if nothing was started; otherwise it stops the memory monitor, then sends SIGKILL to the process group and the leader PID—matching the existing OOM kill path for creack/pty session leaders so JVMs and other subprocesses do not keep running and skew memory limits (e.g. Java OOM scenarios).

^{Reviewed by Cursor Bugbot for commit 3d028cc. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 8cdade5. Configure here.}

[andy1li]

Thanks to @CaseyMcGuire and @mathilch for highlighting the issue!