This PR adds a small standalone helper, cgroupstat, to report CPU usage for the current cgroup using Linux cgroup v2 cpu.stat, without modifying the existing sadc/sar collectors.

Motivation

There is ongoing interest (e.g. issue #260) in being able to scope sysstat-style collection to containers and other cgroup-based workloads. Today this typically requires out-of-band tooling or reading /sys/fs/cgroup/.../cpu.stat directly.
cgroupstat provides a focused, low-risk way to explore and use cgroup-scoped CPU accounting, especially in container environments (e.g. Docker, Kubernetes, CircleCI) where only the container’s cgroup resources are relevant.

Implementation details

• New file: cgroupstat.c (no changes to existing sysstat sources or build system).
• cgroup path detection:
  • Reads /proc/self/cgroup and looks for the cgroup v2 line (prefix 0::).
  • Extracts the cgroup-relative path (e.g. /user.slice/.../app.slice/app-xyz.scope).
  • Builds the corresponding cpu.stat path assuming the unified hierarchy mount at /sys/fs/cgroup:
    • /sys/fs/cgroup<rel_path>/cpu.stat
  • If no 0:: line is found, falls back to /sys/fs/cgroup/cpu.stat.
• Parsing cpu.stat:
  • Reads and stores:
    • usage_usec
    • user_usec
    • system_usec
  • Ignores other fields if present for forward compatibility.
• No changes to:
  • sadc, sar, or activity definitions.
  • Existing test harness (do_test / tests/root*).

CLI usage

• One-shot snapshot of the current cgroup’s CPU usage:

./cgroupstat

Outputs, for example:

cgroup cpu.stat: /sys/fs/cgroup/.../cpu.stat
  usage_usec  = 1390085755
  user_usec   = 986069568
  system_usec = 404016187

• Interval sampling (deltas between samples):

./cgroupstat <interval_seconds> <count>
# e.g.
./cgroupstat 1 2

Sample output:

cgroup cpu.stat delta over 1 s (sample 1/2):
  usage_usec  = 38847
  user_usec   = 27171
  system_usec = 11676
...