feat: add memory usage benchmark implementation

[jim-junior]

This PR adds Memory benchmarking implementations

It measures the Sandbox footprint and also collects the USS, PSS and RSS memeory sets for the shim process associated with a container. The default image used currently in the experiment.yml is the nginx image.

How it works

The implementation is basically a Go script derived from this sequence of shell command steps.

Cgroups Metrics (Sandbox foot print)

1. First it pulls the image with nerdctl, then creates and runs the container.
2. Next it gets the container PID

   PID=$(sudo nerdctl inspect --format '{{.State.Pid}}' $CONTAINER)

3. Then it locates the cgroup

   CGROUP_REL=$(awk -F: '$1 == "0" {print $3}' /proc/$PID/cgroup)
   CGROUP="/sys/fs/cgroup${CGROUP_REL}"

4. Then it collects the cgroups metrics

   sudo cat "$CGROUP/memory.current"
   sudo cat "$CGROUP/memory.stat"
   sudo cat "$CGROUP/memory.peak"

5. Then pparses them and exports them as JSON

Shim process Memory Usage

1. Gets the container PID (this happens when the image is pullled and container is running)

   CID=$(sudo nerdctl inspect --format '{{.Id}}' $CONTAINER)

2. Then get the shim PID associated with that container ID:

   SHIM_PID=$(sudo pgrep -af containerd-shim-runc-v2 | awk -v id="$CID" '$0 ~ id {print $1; exit}')

3. then get the uss, pss, rss

   sudo cat /proc/$SHIM_PID/smaps_rollup

4. Lastly parses the smaps_rollup file contents into a Go string map, and then collects the USS, PSS AND RSS

[cmainas]

Hello @jim-junior ,

the merging of the CPU PR has created some conflicts with this PR.

[cmainas]

Thank you @jim-junior .