Improve fault tolerance of addon delegation requests

[johansenja]

Currently there is no error handling for server_addon/delegate requests - which I think makes sense on the whole (in line with comment Do not wrap this in error handlers. Server add-ons need to have the flexibility to choose if they want to include a response or not as part of error handling, so a blanket approach is not appropriate.).

However, I think there is just one case where I think a bit of error handling might be quite helpful - namely if the requested extension is not available. Currently, @server_addons[name]&.execute(request, params) swallows any typos or mis-registered servers - but IMO it should be an error with a clear message reported.

With the current delegate_request implementation in runner_client, make_request sends the server_addon/delegate message to the server, and then gets stuck waiting for a response (which isn't going to come, because @server_addons[name]&.execute(request, params) is nil) - if I'm understanding things correctly.

      def delegate_request(server_addon_name:, request_name:, **params)
        make_request(
          "server_addon/delegate",
          server_addon_name: server_addon_name,
          request_name: request_name,
          **params,
        )
      rescue MessageError
        nil
      end

I'd be curious for any other opinions on this, and if it's of interest I can add some test cases (and/or any more polish)

[andyw8]

Overall this makes sense to me. But I would skip the DidYouMean checks, as this seems like something that would only happen once during the development of an add-on.

@vinistock should be back next week, so let's hold off for his opinion.