Error case handling

[edwh]

There are several error cases which need review to check we handle them.

Some may be mitigated by the global approach we have for spotting API errors and popping up a toast.

• [x] We should consider whether we need a similar toast approach for JS errors (both SSR and client).
• [x] Check we know what the axios behaviour is for API retries. In the past under high load or server flakiness we have benefited a lot from transparent retry of API calls. We should comment somewhere what the retry strategy is.

Specific error cases to review and perhaps test.

• [ ] yahoologin around line 64 if server login fails
• [ ] OurFilePond around line 142 if file upload fails
• [ ] find/whoami and give/whoami if compose/submit fails
• [ ] auth.js/loginModal when sign up fails - do we get an error?

[edwh]

General approach is that Sentry catches both JS errors and API errors. I've used this to whittle down the number of errors very considerably, and will continue to do so.

There are still some rare but important error cases listed above which need work.