@linode/api-v4: Race condition for setToken() with multiple token

Nobody has claimed this yet.

Hi team,

In general, we call `setToken()` and then an endpoint, like:
```
public async getKubeVersions() {
setToken(linodeToken);
return await getKubernetesVersions();
}
```
The issue is we call the endpoints very often with a lot of different linode tokens. Behind `@linode/api-v4` is a single axios instance, we keep chang it's header for different requests using `setToken()` and this may cause a race condition. For example, if two `getKubeVersions()` call at the same time, we may have:

```
setToken(tokenA);
setToken(tokenB);
getKubernetesVersions(); // mean to use tokenA but actually used tokenB
getKubernetesVersions();
```

Do we have a way to avoid this? For example, to create our own axois instance and call the endpoints from there?

Thanks.

Start by tracing how setToken() updates the shared axios instance used by getKubernetesVersions() in @linode/api-v4. Reproduce concurrent calls with different tokens and determine the supported isolation point. Done means concurrent requests consistently use their intended token, with coverage for the race scenario.

Written by the indexing model from the issue text.