stack pods: use exec_pods from the server for the EXEC column

Problem

Since #188, stack pods works out on the client which pod stack exec picks for each container. The client data has two gaps:

The pod list has no deletion state. A pod that is shutting down, but still Running and Ready, can get the hint. The server skips that pod.
container_states mixes init containers with normal containers. An init container with the name of an exec container can get the hint. The server picks only from the normal containers.

k8s-stack-manager sends exec_pods for each chart with exec hints (omattsson/k8s-stack-manager#767, PR omattsson/k8s-stack-manager#768). It maps each name in exec_containers to the pod that the exec route picks for -c <name>.

Add ExecPods map[string]string (exec_pods,omitempty) to types.ChartStatus.
In execColumn (cli/cmd/stack_inspect.go): when exec_pods is present, put each valid container name on the row of the pod that exec_pods names. Keep the * on the default container.
Without exec_pods (older server), keep the current client rule.
A container whose pod is not in the listed pods goes on the line under the table, as now.
JSON and YAML output carry exec_pods as received.
Update the help text and docs/commands.md: the known limits apply only to older servers.

Table-driven cases with exec_pods: a terminating pod sorted first, an init container with the same name, a pod name not in the list, and an invalid container name in the map.
The fallback without exec_pods stays as it is.

omattsson/k8s-stack-manager#768 must be merged and released first.