Ignore diacritics in autocomplete

Works for:
- Choice
- Choice List
- Reference
- Reference List

Co-Authored-By: Louis Delbosc <louis.delbosc.prestataire@anct.gouv.fr>

app/client/lib/ACIndex.ts

@@ -7,7 +7,7 @@
  * "lush" would only match the "L" in "Lavender".
  */
 
-import {localeCompare, nativeCompare, sortedIndex} from 'app/common/gutil';
+import {localeCompare, nativeCompare, removeDiacritics, sortedIndex} from 'app/common/gutil';
 import {DomContents} from 'grainjs';
 import escapeRegExp = require("lodash/escapeRegExp");
 
@@ -17,6 +17,12 @@ export interface ACItem {
   cleanText: string;
 }
 
+// Returns a normalized, trimmed, lowercase version of a string, so that autocomplete is case-
+// and accent-insensitive.
+export function cleanText(text: string): string {
+  return removeDiacritics(text).trim().toLowerCase();
+}
+
 // Regexp used to split text into words; includes nearly all punctuation. This means that
 // "foo-bar" may be searched by "bar", but it's impossible to search for punctuation itself (e.g.
 // "a-b" and "a+b" are not distinguished). (It's easy to exclude unicode punctuation too if the
@@ -91,7 +97,7 @@ export class ACIndexImpl<Item extends ACItem> implements ACIndex<Item> {
   // The main search function. SearchText will be cleaned (trimmed and lowercased) at the start.
   // Empty search text returns the first N items in the search universe.
   public search(searchText: string): ACResults<Item> {
-    const cleanedSearchText = searchText.trim().toLowerCase();
+    const cleanedSearchText = cleanText(searchText);
     const searchWords = cleanedSearchText.split(wordSepRegexp).filter(w => w);
 
     // Maps item index in _allItems to its score.