Inconsistency between `juju bootstrap` and `juju deploy` when dealing with Kubernetes configuration flags

[Jonatas Baldin]

The `juju boostrap` and `juju deploy` commands have inconsistencies in their acceptable Kubernetes config options.

You can do (`loadbalancer` in lower):
```
$ juju bootstrap --config controller-service-type=loadbalancer my-k8s
```

But in `juju deploy`:
```bash
$ juju deploy cs:~mattermost-charmers/mattermost --config kubernetes-service-type=loadbalancer
```

Generates this error:
```bash
$ juju status
Model Controller Cloud/Region Version SLA Timestamp
my-model my-k8s my-k8s 2.9-rc6 unsupported 16:45:32+01:00

App Version Status Scale Charm Store Rev OS Address Message
mattermost mattermost:5.31.0 error 1 mattermost charmstore 17 kubernetes creating or updating service for mattermost: Service "mattermost" is invalid: spec.type: Unsupported value: "loadbalancer": supported values: "ClusterIP", "ExternalName", "LoadBalancer", "NodePort"
```

There are maybe more inconsistencies that I didn't look around, but as a user I'd expect everything to be lowercase or not.

[Ian Booth (wallyworld)]

Both the charm and bootstrap use the lower case version "loadbalancer".
deploy should be made consistent with this.
We should probably accept both "LoadBalancer" and "loadbalancer" and convert to "loadbalancer"

[Ian Booth (wallyworld)]

https://github.com/juju/juju/pull/12644 fixes the issue
A validation check is put in the ensure the supplied service type is correct.

We support "loadbalancer", "external", "cluster" to match what's used for bootstrap and in charm metadata. As a fail safe, any already deployed charms which use "LoadBalancer" or "ClusterIP" etc will continue to work.