feat(tool): support non-utf8 file encodings - #49881

Issue for this PR

Closes #45924

Type of change

• New feature
• Bug fix
• Refactor / code improvement
• Documentation

What does this PR do?

edit, write, apply_patch and the patch engine decoded every file as UTF-8 with a non-fatal decoder, so a GBK/Shift-JIS/Big5 file turned into U+FFFD garbage and the corruption was written back permanently.

This adds per-read encoding detection (BOM for UTF-8/UTF-16LE/UTF-16BE, then UTF-8 validity, then a new configurable fallback file_encoding in opencode.json, default utf-8 so default behavior is unchanged) in a shared util/encoding module, wired into edit/write/read/apply_patch and the patch engine. Files are written back in the encoding they were loaded with, and new files are created in the fallback encoding. Non-UTF-8 decoding goes through iconv-lite, lazily imported only when needed.

Related: #45925 rejects non-UTF-8 files in the v2 core tools. This PR is complementary: it makes the v1 tools encoding-aware so legacy files stay editable.

How did you verify your code works?

• New test/tool/encoding.test.ts (13 tests): GBK edit/write/read round-trips verified at the byte level, UTF-16LE/BE BOM round-trips, UTF-8 BOM preserved, file left byte-identical when an edit does not match.
• bun test test/tool/ test/patch/ from packages/opencode — no new failures (the 4 Windows path-normalization failures also fail on vanilla dev).
• bun typecheck passes for packages/opencode and packages/core.

Screenshots / recordings

Not a UI change.

Checklist

• I have tested my changes locally
• I have not included unrelated changes in this PR