UTF-8 validation in CLI

`juju config` employs UTF-8 validation on provided values, which was added in
https://github.com/juju/juju/pull/150
I believe the original reason for this was as follows. `juju config` supports reading configuration values from a file using an `@`. For example,
```
$ juju config <email address hidden>
```
will set the value for `key1` to the contents of `foo.txt`. In this case, it makes sense to check that the contents of `foo.txt` are valid UTF-8 before you set `key1`.

None of the other config commands use UTF-8 validation, and it seems to be used only sporadically throughout the CLI. Commands vary in whether they accept invalid UTF-8 sequences - some just accept it:
```
$ juju model-config ftp-proxy=$(printf %b '\xc3\x28')
$
```
```
$ juju deploy $(printf %b '\xc3\x28')
ERROR no charm was found at "\xc3("
```
```
$ juju bootstrap lxd $(printf %b '\xc3\x28')
Creating Juju controller "�(" on lxd/localhost
...
```

while some will complain:
```
$ juju config hello-juju port=$(printf %b '\xc3\x28')
ERROR value for option "port" contains non-UTF-8 sequences
```
```
$ juju add-model $(printf %b '\xc3\x28')
ERROR "\xc3(" is not a valid name: ...
```

It's worth thinking about whether this should be done more consistently:
 - Do we even need UTF-8 validation at all in `config`, `add-model`, etc?
 - Should we do this consistently across all the config commands, or even all commands?
i.e. when is UTF-8 validation appropriate/necessary, and when is it not needed?