docs: translate Japanese comments to English

[hashiiiii]

Summary

Translate every in-code Japanese comment and description string to English across the whole repository, so the code language matches the already-English docs (README, SECURITY.md). Comments and config descriptions only — no logic changes.

Motivation

Closes #55.

The user-facing docs are English while in-code comments were mostly Japanese. That (a) makes intent hard to read for non-Japanese contributors and (b) duplicates the language split between docs and code, raising maintenance cognitive load. Unifying on English prepares the repo for outside contribution and OSS visibility.

Changes

• Translated Japanese comments to English in core/ (14 files), cli/ (7), extension/ (34), and editor/ (8), plus build.zig comments and the renovate.json description fields.
• Also covered hidden config files that the issue's grep command misses (ripgrep skips dot-directories unless --hidden is passed): .gitattributes, .github/workflows/ci.yml, .github/workflows/release.yml.
• cli/src/mcp.zig: the "truncateTree never splits a multibyte utf-8 character" test used the intentional literal "あああ". Replaced it with "€€€" — U+20AC is also exactly 3 bytes in UTF-8, so the input byte layout and every assertion (len = 50008, cut at byte 49999) are unchanged and the test still exercises a mid-character cut. Surrounding comments now reference '€'. This keeps the repo fully CJK-free without weakening the regression test.
• Verified via diff analysis that only comment text changed: every non-comment-marker line in the diff is a code; // comment pair whose code portion is byte-identical between the removed and added side.

While translating, I corrected one apparent typo in .gitattributes: CSRF -> CRLF. The surrounding comment is about preventing automatic line-ending conversion on Windows checkout, where CSRF (a security term) is meaningless; the intent is clearly CRLF. Flagging it explicitly since this issue is scoped to translation.

Testing

Reproduced the CI matrix locally — all green:

• CJK grep repo-wide including hidden dirs: rg --hidden '[\x{3040}-\x{309F}\x{30A0}-\x{30FF}\x{4E00}-\x{9FFF}\x{3000}-\x{303F}\x{FF00}-\x{FFEF}]' -> 0 hits
• core: zig build lint, zig build test, zig build perf -> all pass (perf: 50000 objects in 174 ms, ceiling 600 ms)
• extension: zig build wasm, node --test core/tests/*.test.mjs (7 pass), npm run size|lint|typecheck|test|build -> all pass (vitest 142 pass, 13 files)
• editor: dotnet csharpier check . --no-msbuild-check -> pass (8 files)
• Extension E2E (Playwright) was not run locally (comment-only change; typecheck and build pass) and is left to CI.

Follow-up candidates (out of scope here — translation only, not fixed):

• cli/src/main.zig:494 — stale Task 9 reference (was already English).
• editor/Editor/Cli.cs:131 — comment ties the absence of SHA256SUMS to "a release before v0.2.0"; worth verifying against the actual release history.