Don't consider textual characters to be emoji (#12582)

* Don't consider textual characters to be emoji

We were using emojibase-regex to match emoji within messages. However, the docs (https://emojibase.dev/docs/regex/) state that this regex matches both emoji and text presentation characters. This is not what we want, and will result in false positives for characters like '↔' that could turn into an emoji if paired with a variation selector. Unfortunately, none of the other regexes provided by Emojibase do what we want either (https://github.com/milesj/emojibase/issues/174). In the meantime, browser support for the RGI_Emoji character sequence class has made it feasible to write an emoji regex by hand, so that's what I've done.

* Add a fallback for BIGEMOJI_REGEX as well

test/HtmlUtils-test.tsx

@@ -107,6 +107,12 @@ describe("bodyToHtml", () => {
         expect(html).toMatchInlineSnapshot(`"<span class="mx_EventTile_searchHighlight">test</span> foo &lt;b&gt;bar"`);
     });
 
+    it("generates big emoji for emoji made of multiple characters", () => {
+        const { asFragment } = render(bodyToHtml({ body: "👨‍👩‍👧‍👦 ↔️ 🇮🇸", msgtype: "m.text" }, [], {}) as ReactElement);
+
+        expect(asFragment()).toMatchSnapshot();
+    });
+
     it("should generate big emoji for an emoji-only reply to a message", () => {
         const { asFragment } = render(
             bodyToHtml(
@@ -132,6 +138,12 @@ describe("bodyToHtml", () => {
         expect(asFragment()).toMatchSnapshot();
     });
 
+    it("does not mistake characters in text presentation mode for emoji", () => {
+        const { asFragment } = render(bodyToHtml({ body: "↔ ❗︎", msgtype: "m.text" }, [], {}) as ReactElement);
+
+        expect(asFragment()).toMatchSnapshot();
+    });
+
     describe("feature_latex_maths", () => {
         beforeEach(() => {
             jest.spyOn(SettingsStore, "getValue").mockImplementation((feature) => feature === "feature_latex_maths");