fix(kernel): interrupt IPython subprocess groups

[lucasrgt]

Summary

Fixes interrupted IPython cells that stay busy until a subprocess naturally exits.

Prime Agent currently launches direct and forked kernels inside the daemon process group. ipykernel deliberately forwards interrupt_request to descendants only when the kernel is its process-group leader; otherwise it signals the kernel PID alone. Commands such as os.system()/subprocess.wait() can therefore leave their child running while the kernel stays occupied for minutes.

Changes

• launch direct Unix kernels with detached: true so each kernel owns a session/process group
• call setsid() in forkserver kernel children before starting ipykernel
• signal the isolated process group during dispose/restart and abandoned-fork cleanup, preventing orphan descendants
• add real-kernel regression coverage for both direct and forkserver launch paths

Windows retains the existing behavior.

Validation

• npm run check
• focused goal/kernel tests: 53/53 passed
• new Linux real-kernel regression: direct and forkserver paths both pass; after abort, the next cell executes immediately

The regression test launches a sleeping subprocess, aborts the cell, verifies kernel PID == PGID, and proves a follow-up cell completes instead of raising KernelBusyAfterInterruptError.

Note

Fix interrupt handling in IPython kernels to signal full process groups

• On non-Windows, directly spawned kernels are started with detached: true so they become process-group leaders; forked kernels call os.setsid() for the same effect.
• Termination and interrupt signals now target the kernel's Unix process group (negative PID) rather than just the kernel PID, ensuring child processes spawned during cell execution are also interrupted.
• Adds killIsolatedKernelProcessGroup in index.ts and terminateForkedKernel in fork-server.ts with fallback to direct PID signaling when the group is not yet established.
• Adds integration tests in kernel-process-group.test.ts (skipped on Windows) that verify interrupts reach descendant processes and the kernel remains usable afterward.
• Behavioral Change: kernels that previously ignored interrupts or left child processes running after abort will now receive SIGTERM/SIGINT across the whole process group.

^{Macroscope summarized 8dcfff8.}

[sethkarten]

Thank you for the report and proposed work. This root cause is now covered by maintainer-owned stacked PR #1162, authored independently from upstream/main.

We did not inspect or reuse this PR's diff, branch, commits, implementation code, or tests; its public description/comments were used only as a bug report. To keep one review surface, this PR is superseded by #1162 and is being closed.

The complete review stack is #1158–#1165. It is being left unmerged for human review after CI and review-bot findings are cleared.