fix: properly decode UTF-8 from GitHub API base64 content

[mkobayashime]

Summary

• Fix mojibake (character corruption) in manifest.json name fields when syncing bookmarks from GitHub repositories
• Properly decode UTF-8 encoded content from GitHub's base64 API responses

Problem

The GitHub API returns file content as base64-encoded strings. When using atob() directly to decode, multi-byte UTF-8 characters (like Japanese text) were incorrectly decoded as individual Latin-1 characters, causing garbled text in bookmark names.

Solution

Updated fetchFileContent in lib/github/api.ts to properly decode UTF-8:

1. Decode base64 to binary string via atob()
2. Convert binary string to Uint8Array of bytes
3. Decode bytes as UTF-8 using TextDecoder()

Test plan

• Create a repository with a manifest.json containing Japanese (or other multi-byte UTF-8) characters in the name field
• Connect the repository in GitMarks
• Sync bookmarks and verify that names display correctly without mojibake