fix: clean up VM runners on cluster destroy

This never worked properly, as `Wait()` doesn't work for child
processes, and `talosctl cluster destroy` is not a child of processes
created by `talosctl cluster create`.

Signed-off-by: Andrey Smirnov <andrey.smirnov@siderolabs.com>

pkg/provision/providers/vm/process.go

@@ -5,10 +5,12 @@
 package vm
 
 import (
-	"errors"
 	"fmt"
 	"os"
 	"syscall"
+	"time"
+
+	"github.com/siderolabs/go-retry/retry"
 )
 
 // StopProcessByPidfile stops a process by reading its PID from a file.
@@ -43,13 +45,13 @@ func StopProcessByPidfile(pidPath string) error {
 		return fmt.Errorf("error sending SIGTERM to %d (path %q): %w", pid, pidPath, err)
 	}
 
-	if _, err = proc.Wait(); err != nil {
-		if errors.Is(err, syscall.ECHILD) {
-			return nil
+	// wait for the process to exit, this is using (unreliable and slow) polling
+	return retry.Constant(30*time.Second, retry.WithUnits(100*time.Millisecond)).Retry(func() error {
+		err = proc.Signal(syscall.Signal(0))
+		if err == nil {
+			return retry.ExpectedErrorf("process %d still running", pid)
 		}
 
-		return fmt.Errorf("error waiting for %d to exit (path %q): %w", pid, pidPath, err)
-	}
-
-	return nil
+		return nil
+	})
 }