common http: set a low timeout on requests

We should be default set quite a low timeout on requests, maybe 20 seconds.

This means that if the backend service is unavailable, you're more likely to get a "request timed out" adaptor error than for the whole job to be killed without any feedback at all.

We've seen problems today where the backend service returns a 502 errror after a minute or something, but the run never sees that because it was killed before. Resulting in a frustrating user experience.

In the event of a http timeout, the error message should say something like "set the timeout option to change the timeout duration`. Although if we do that in common, we don't know what the actual adaptor's API is. So we need to think about that - the point is to nudge the user to making a simple fix if they have long requests.

No file or test is named. Inspect the common HTTP request path and how adaptor APIs expose options, then reproduce an unavailable backend or delayed 502 response. Done means requests use a low default timeout, timeout failures provide actionable feedback, and longer requests can override it.