Better propagation of dry-run tx errors

[sug0]

We often see this pattern in Namada code:

            if request.path == RPC.shell().dry_run_tx_path() {
                super::dry_run_tx(ctx, &request)
            } else {
                self.rpc.handle(ctx, &request)
            }

This is because dry-run tx errors are poorly propagated to the caller. We should come up with a mechanism to better propagate errors.