Remove spreadString in favor of polyfilled Array.from

Fix https://gitlab.com/gitlab-org/gitlab-ce/merge_requests/10120

We use this to split up astral symbols

```js
// ["🖐", "🏿"]
Array.from('🖐🏿');
```

app/assets/javascripts/behaviors/gl_emoji/is_emoji_unicode_supported.js

-import spreadString from './spread_string';
-
 // On Windows, flags render as two-letter country codes, see http://emojipedia.org/flags/
 const flagACodePoint = 127462; // parseInt('1F1E6', 16)
 const flagZCodePoint = 127487; // parseInt('1F1FF', 16)
@@ -20,7 +18,7 @@ function isKeycapEmoji(emojiUnicode) {
 const tone1 = 127995;// parseInt('1F3FB', 16)
 const tone5 = 127999;// parseInt('1F3FF', 16)
 function isSkinToneComboEmoji(emojiUnicode) {
-  return emojiUnicode.length > 2 && spreadString(emojiUnicode).some((char) => {
+  return emojiUnicode.length > 2 && Array.from(emojiUnicode).some((char) => {
     const cp = char.codePointAt(0);
     return cp >= tone1 && cp <= tone5;
   });
@@ -30,7 +28,7 @@ function isSkinToneComboEmoji(emojiUnicode) {
 // doesn't support the skin tone versions of horse racing
 const horseRacingCodePoint = 127943;// parseInt('1F3C7', 16)
 function isHorceRacingSkinToneComboEmoji(emojiUnicode) {
-  return spreadString(emojiUnicode)[0].codePointAt(0) === horseRacingCodePoint &&
+  return Array.from(emojiUnicode)[0].codePointAt(0) === horseRacingCodePoint &&
     isSkinToneComboEmoji(emojiUnicode);
 }
 
@@ -42,7 +40,7 @@ const personEndCodePoint = 128105; // parseInt('1F469', 16)
 function isPersonZwjEmoji(emojiUnicode) {
   let hasPersonEmoji = false;
   let hasZwj = false;
-  spreadString(emojiUnicode).forEach((character) => {
+  Array.from(emojiUnicode).forEach((character) => {
     const cp = character.codePointAt(0);
     if (cp === zwj) {
       hasZwj = true;

app/assets/javascripts/behaviors/gl_emoji/spread_string.js

-// https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/String/charCodeAt#Fixing_charCodeAt()_to_handle_non-Basic-Multilingual-Plane_characters_if_their_presence_earlier_in_the_string_is_known
-function knownCharCodeAt(givenString, index) {
-  const str = `${givenString}`;
-  const end = str.length;
-
-  const surrogatePairs = /[\uD800-\uDBFF][\uDC00-\uDFFF]/g;
-  let idx = index;
-  while ((surrogatePairs.exec(str)) != null) {
-    const li = surrogatePairs.lastIndex;
-    if (li - 2 < idx) {
-      idx += 1;
-    } else {
-      break;
-    }
-  }
-
-  if (idx >= end || idx < 0) {
-    return NaN;
-  }
-
-  const code = str.charCodeAt(idx);
-
-  let high;
-  let low;
-  if (code >= 0xD800 && code <= 0xDBFF) {
-    high = code;
-    low = str.charCodeAt(idx + 1);
-    // Go one further, since one of the "characters" is part of a surrogate pair
-    return ((high - 0xD800) * 0x400) + (low - 0xDC00) + 0x10000;
-  }
-  return code;
-}
-
-// See http://stackoverflow.com/a/38901550/796832
-// ES5/PhantomJS compatible version of spreading a string
-//
-// [...'foo'] -> ['f', 'o', 'o']
-// [...'🖐🏿'] -> ['🖐', '🏿']
-function spreadString(str) {
-  const arr = [];
-  let i = 0;
-  while (!isNaN(knownCharCodeAt(str, i))) {
-    const codePoint = knownCharCodeAt(str, i);
-    arr.push(String.fromCodePoint(codePoint));
-    i += 1;
-  }
-  return arr;
-}
-
-export default spreadString;