Fix Error Handling

[confused-Techie]

It was brought to my attention on Discord that a61aa7e121bc93723f675198778b6006fdd66738 broke error handling in PPM.

The clue to find out what had gone wrong was the fact that the unstar command still worked.

This is because in that PR I had told superagent a list of status codes to not error on. This means that when we got, say a 404, the actual error object returned to the callback would be null even though those commands then relied on the error object to determine what had gone wrong.

So what I've done, is provide both the error and body to the request.getErrorMessage() so that we can check for errors in either of them.

Meaning when a request fails we will:

1. Look first for within the superagent error object for err.response.body then for err.response.error then for err
2. If that fails, we can then look at the body for body.message then body.error then finally body

This helpfully means we prioritize any errors returned by superagent then any errors returned by the backend. And if all else fails, we just tell the user whatever message was returned to us by the backend, which can be helpful in case the backend's error schema ever changes by accident or on purpose and isn't updated here. So that way we can always inform the user exactly what the backend is thinking.

[confused-Techie]

We are seeing more failures of NodeJS 18 here, like we are elsewhere. So those shouldn't be considered during a final review of the PR.

[confused-Techie]

Thanks a ton for the approval @DeeDeeG! I'll go ahead and merge this one