[receiver/k8sclusterreceiver] Report entity state periodically to be immune to event losses and reordering

[tigrannajaryan]

Entity state is represented as log records and delivered via regular log pipeline and exporter queue. The events in the queue may be easily reordered. Events may be also lost if queue is overflowed.

To ensure entity's state is eventually correctly delivered to the destination the design requires that entity state event is periodically reported even if the entity does not change.

This requires k8sclusterreceiver to perform periodic collection (in addition to on-change) of K8s metadata.

Note that it is a rare situation for event reordering to result in incorrect state at the destination. The same entity has to change twice quickly enough to end up in the same or immediately adjacent batch of events, then batches end up being reordered in the exporter queue (or if in the same batch, the batch is split and then reordered). This is possible but is going to be quite rare so we are ok that when this happens it takes a few minutes to be corrected.

Because of this I think a reasonable default interval is probably around 5-10min. This should ensure that the increased load from repetition is reasonably small yet any errors are corrected quickly enough.