Attribute accessibility improvements, and icons

[cerebrate]

I made some modifications to this integration for my own use, so thought I'd share them back in case you think they're worth including:

• For each type of entity, I broke some (the ones I most needed, in this case, for tracking status) of the k8s properties out into top-level attributes to make them easier to access with, for instance, the auto-entities card, and automations without having to repeat templates each time.
• I moved the complete k8s object dump under the "raw" attribute, rather than leaving it to several top-level attributes; this to ensure the integration controls the attribute namespace and future k8s changes don't step on anything. This is a breaking change, but I think a worthwhile one?
• I added an "ok" attribute checking the status of each type of entity: that nodes are KubeletReady, that deployments and daemonsets have the correct number of pods available, and that pods are in the running state; I then also use this to provide provide icons which change with ok/not-ok status for each type of entity.

Small stuff, but hopefully worthwhile!

[Adi146]

Hey, thanks for your Pull Request. I like changes with the raw attribute and the icon but I'm not sure if we really need the ok attribute. The node and pod sensor basically already represent these in the state of the sensor. Only the daemonset and deployment sensor could use this attribute but I think we should rather also display this in the state. What do you think about this?

[cerebrate]

Well, I'm not wedded to the idea of the ok attribute. I added it mostly because I already needed to write a simple boolean is-everything-okay check to power the icon, and having done that, thought I might as well expose it. For basic automations (such as my "hey, this is broken, go check on it" announcement, having a simple true/false check is handy, and means I don't have to remember when writing them if it's "KubeletReady" or "Ready" or "Running", or whatever.

(I didn't put it in the state because it seemed like a breaking change that wasn't needed; if people are already relying on the state being the number of ready pods for daemonset/deployment, or the text state in node and pod, I didn't want to break their automation.)

But like I said, I'm not wedded to the idea. If you think it'd be better in the state, I have no issue with that. I'd just like to have that bool available somewhere.

[Adi146]

Yeah, I think we should rather use this check in the state for daemonsets and deployments. For Nodes and Pods we should keep the sensor state as it is right now, this way it is more flexible to use. I don't care about breaking changes at this point, because as far as I know are we the only two using this at the moment. What do you think about that? Could you please adjust your Pull Request to that?

[cerebrate]

Sounds good to me. I've made the changes.