Make decoding base64 in ContentData much faster

When exporting a large trace from gemini.google.com (~5s recording), it
was taking ~3.4s. Now it takes ~0.5s.

The expensive part was this:

> const bytes = Uint8Array.from(binaryString, m => m.codePointAt(0) as
number);

I believe the slowness came from:

1. overhead of v8 calling the mapper function for each character of the
   base64 string
2. needing to create a string instance of length 1 for each character

Additionally: now using charCodeAt instead of codePointAt. I didn't
notice a measurable improvement from this change, but we don't need the
complexity of code points when processing base64.

Bug: 444483828
Change-Id: I5c6ab8a345637672277062e84f4fdf05a06bdbbe
Reviewed-on: https://chromium-review.googlesource.com/c/devtools/devtools-frontend/+/6940949
Commit-Queue: Connor Clark <[email protected]>
Auto-Submit: Connor Clark <[email protected]>
Reviewed-by: Paul Irish <[email protected]>

Author: Connor Clark

front_end/models/text_utils/ContentData.ts

@@ -76,7 +76,12 @@ export class ContentData {
     }
 
     const binaryString = window.atob(this.#contentAsBase64 as string);
-    const bytes = Uint8Array.from(binaryString, m => m.codePointAt(0) as number);
+    const len = binaryString.length;
+    const bytes = new Uint8Array(len);
+    for (let i = 0; i < len; i++) {
+      bytes[i] = binaryString.charCodeAt(i);
+    }
+
     this.#contentAsText = new TextDecoder(this.charset).decode(bytes);
     return this.#contentAsText;
   }