[SidecarContainer] Need to provide sidecar implementation best practices

[SergeyKanzhelev]

This is a follow up from kubernetes/enhancements#753 (comment)

We need to improve the page https://kubernetes.io/docs/tutorials/configuration/pod-sidecar-containers/ with the best practices on how to implement sidecar containers. Things like exiting with 0 on SIGTERM

[SergeyKanzhelev]

/sig node
/kind feature

[sftim]

/priority backlog
/language en
/triage accepted

[pegasas]

/assign @pegasas

[SergeyKanzhelev]

Some interesting issues that we need to cover:

1. If a sidecar never had a chance to execute the startup probe, it will prevent the main container from even starting: SideCar startupProbe on 1.28 kubernetes#120848
2. Sidecar implemented as a built-in type always starts before the container. So it will block the main container start before the sidecar's image is pulled. See [Sidecar Containers] Ability to start sidecar and main containers in parallel kubernetes#126471 The best practice here is to postpone all heavy operations to as late as possible.