Add use <tunit> instead of <span> for Deepl translation

To avoid getting results like "... bla.&quot;</span>" ->
"... bla</span>.&quot;" from Deepl we need to set set splitting_tags
option to an element name (that's not as common as span).

lib/outlook/translators.ex

@@ -73,7 +73,7 @@ defmodule Outlook.Translators do
 
   defp prepare_article(tree) do
     # Logger.info "so far."
-    HtmlDocComponent.render_doc(%{tree: tree})
+    HtmlDocComponent.render_doc(%{tree: tree, tunit_tag: "tunit"})
     |> Phoenix.HTML.Safe.to_iodata()
     |> IO.iodata_to_binary()
   end
@@ -86,7 +86,7 @@ defmodule Outlook.Translators do
   def process_translation(translation, tunit_ids) do
     tunit_map = translation
     |> Floki.parse_fragment!
-    |> Floki.find("span.tunit")
+    |> Floki.find("tunit")
     |> Enum.map(fn {_,atts,cont} ->
       %TranslationUnit{
         nid: Enum.find(atts, fn {k,_} -> k == "nid" end) |> Tuple.to_list |> Enum.at(1),

lib/outlook/translators/deepl.ex

@@ -26,6 +26,8 @@ defmodule Outlook.Translators.Deepl do
       [
         {"source_lang", options.source_lang},
         {"target_lang", options.target_lang},
+        {"tag_handling", "xml"},
+        {"splitting_tags", "tunit"},
         {"file", content, {"form-data", [{:name, "file"}, {:filename, "datei.html"}]}, []}
       ]
     )