Merge branch 'cache-autocomplete-results' into 'master'

Cache autocomplete results

## What does this MR do?

Caches the results of the autocomplete AJAX call to stop new GFM inputs on the page from requesting new data. The cache is then cleared on each new page so that it doesn't stay around per project.

See merge request !5043

app/assets/javascripts/gfm_auto_complete.js.coffee

@@ -4,7 +4,7 @@ window.GitLab ?= {}
 GitLab.GfmAutoComplete =
   dataLoading: false
   dataLoaded: false
-
+  cachedData: {}
   dataSource: ''
   # Emoji
@@ -55,7 +55,7 @@ GitLab.GfmAutoComplete =
     @setupAtWho()
     if @dataSource
-      if !@dataLoading
+      if not @dataLoading and not @cachedData
         @dataLoading = true
         # We should wait until initializations are done
@@ -70,6 +70,8 @@ GitLab.GfmAutoComplete =
             @loadData(data)
         , 1000)
+      if @cachedData?
+        @loadData(@cachedData)
   setupAtWho: ->
     # Emoji
@@ -205,6 +207,7 @@ GitLab.GfmAutoComplete =
     $.getJSON(dataSource)
   loadData: (data) ->
+    @cachedData = data
     @dataLoaded = true
     # load members

app/views/layouts/_init_auto_complete.html.haml

@@ -3,4 +3,5 @@
 - if @noteable
   :javascript
     GitLab.GfmAutoComplete.dataSource = "#{autocomplete_sources_namespace_project_path(project.namespace, project, type: @noteable.class, type_id: params[:id])}"
+    GitLab.GfmAutoComplete.cachedData = undefined;
     GitLab.GfmAutoComplete.setup();