Let sync op failures use stacktrace viewer

This lets sync ops use the stacktrace viewer, just like how async ops
currently do. Requires some changes to the CIDER-nREPL code as well
which were submitted as CIDER-nREPL PR 327
clojure-emacs/cider-nrepl#327.

CHANGELOG.md

@@ -4,6 +4,8 @@
 
 ### Bugs fixed
 
+
+* [#1420](https://github.com/clojure-emacs/cider/issues/1420): Show stacktrace buffers for sync requests errors.
 * [cider-nrepl#329](https://github.com/clojure-emacs/cider-nrepl/pull/329): Fix error instrumenting functions that call clojure.tools.logging.
 * [#1632](https://github.com/clojure-emacs/cider/pull/1632): Redefining a function correctly updates eldoc.
 * [#1630](https://github.com/clojure-emacs/cider/pull/1630): The debugger no longer gets confused inside `@` redefs.

nrepl-client.el

@@ -968,11 +968,7 @@ sign of user input, so as not to hang the interface."
         ;; so we have to handle them differently until this is resolved
         (if (member "eval-error" status)
             (funcall nrepl-err-handler)
-          ;; dump the stacktrace in the REPL
-          ;; TODO: This has to be replaced with rendering of the
-          ;; standard stacktrace buffer
-          (cider-repl-emit-interactive-stderr err)
-          (switch-to-buffer-other-window connection)))
+          (cider--render-stacktrace-causes (nrepl-dict-get response "pp-stacktrace"))))
       (when-let ((id (nrepl-dict-get response "id")))
         (with-current-buffer connection
           (nrepl--mark-id-completed id)))