Should `LiveView::{mount, update}` be fallible?

[davidpdrsn]

Currently LiveView::{mount, update} allows you to return an error. If that happens the view is shutdown and the client will reconnect and a new view is created. However I'm not sure there actually is much utility to that. Ideally you should always handle all errors and show them nicely to the user.

So maybe we should consider making those methods not return a result and thus require users to handle all errors.