Fix offer letter filename for accented names

[Shashankojha-ai]

Fixes #87

What
Normalize names with NFKD and strip combining marks before creating offer-letter PDF filenames. This preserves accented Latin characters as ASCII and keeps the member fallback for blank or symbol-only names.

How I tested

• npm run lint — passed (existing warnings only).
• npm test — passed, 124 tests.
• npm run build — passed.
• Added regression tests for plain names, accented names, and blank/symbol-only input. This is a filename-only change, so no visible UI changed and no screenshot is applicable.

[vercel]

Someone is attempting to deploy a commit to the Geetansh Goyal's projects Team on Vercel.

A member of the Team first needs to authorize it.

[github-actions]

✅ Thanks @Shashankojha-ai — the description looks good now. This PR is no longer at risk of being closed.

[github-actions]

📸 How this looks

Built from this PR, so you can see the change without checking it out.

/

Desktop (1280px), top 3000px

Mobile (390px), top 3000px

/learn/open-source

Desktop (1280px), top 3000px

Mobile (390px), top 3000px

_{Screenshots are taken from a production build of this PR, signed out. Pages behind sign-in show the signed-out view. Build log}

[MehulNegi]

Checked out the branch and ran it. npx vitest run lib/offer-letter.test.ts passes 3/3, and npm test passes 124/124. The accented case from #87 works now (Ádítí Sharma -> Aditi-Sharma), and keeping this separate from slugify makes sense since that one lowercases and cuts at 60 chars.

Question: why does this add server-only to package.json? lib/offer-letter.ts already imported it on main without it being listed, and the new test imports lib/filename.ts, which doesn't pull it in. I may be missing something about how Next resolves it, but it looks unrelated to the filename change.

[MehulNegi]

This only catches letters that NFKD splits into base + combining mark. Ones it doesn't split still get dropped. I ran a few through sanitizeFilenameSlug:

Łukasz Nowak      -> ukasz-Nowak
Søren Kierkegaard -> S-ren-Kierkegaard
Đặng Văn Lâm      -> ang-Van-Lam
Weiß              -> Wei

That's the same mangling #87 describes, just for different letters. A small map before the [^a-zA-Z0-9] replace (ł->l, ø->o, đ->d, ß->ss, æ->ae, œ->oe) would cover the common ones, plus a test for one of them. If that's out of scope for this issue, a test that pins the current output would at least make the limit visible.