i18n(zh-CN): Profile

[chongfenglaosiji]

i18n(zh-CN): Profile

What & why

Next batch of the zh-CN catalog: Profile, as a data-only change.

Data-only: web/i18n/zh-CN.json is the only file touched. No Go, template or
JavaScript edits.

Scope

34 new keys:

Prefix	Keys	Covers
profile.msg.*	11	The server-built confirmations and errors, now looked up in the reader's language
profile.account.*	7	Account info block
profile.totp.*	7	Two-factor section
profile.password.*	5	Change-password form
profile.* (shared)	4	Page heading and intro

Catalog goes from 63 shipped keys to 97 keys (31% of English). The
rest stays English by design, so nothing breaks.

Merge order (please read)

This diff is a snapshot of the whole file, so besides its own 34 keys it
contains the settings.default_language.auto wording fix that #131 also carries
(否则用英文 → 否则显示英文). Every batch PR sets that line to the same value, so they
merge cleanly in any order and none of them can silently revert it.

Checklist

• go build ./... compiles
• go test ./internal/i18n/ ./internal/server/ -run 'Translation|Catalog|Language|Locale' passes
• Commits are clean and the PR is focused on one concern
• Data-only, matching docs/TRANSLATING.md
• CHANGELOG.md / version bump — left to you, as before
• Licensed under Apache 2.0

Notes for reviewers

• profile.totp.enabled_hint and profile.totp.not_configured_hint are sentence fragments that follow a status word in the template (…已启用 + 你的账号已通过身份验证器应用保护。), so they are phrased to continue that sentence rather than stand alone.
• profile.account.read_only is （read-only）→ （只读）with the full-width brackets the template does not supply — the English value carries its own brackets, so the Chinese does too.

Verification — run against this exact file, in a Linux container (WSL):

go test ./internal/i18n/ ./internal/server/ -run 'Translation|Catalog|Language|Locale' -count=1
go build ./...

exit=0 — including TestEveryCatalogMatchesEnglishKeysAndPlaceholders and
TestEveryTranslationKeyUsedExistsInEnglish. Same as #131.

One note if you ever test on Windows: a plain checkout there fails
internal/caddy's TestBuildProxyRouteGolden, because that test compares bytes and
core.autocrlf makes the testdata/*.json golden files CRLF on disk while the
generated JSON is LF. It reproduces on an untouched main, so it is unrelated to this
PR — it is just why the run above is documented as Linux.

The pr4 log of that run is kept locally if you want it pasted here.

Catalog facts

• 97/97 keys valid; no key English does not have; no duplicate keys
• Placeholders identical to English on every key
• No empty values; UTF-8 without BOM; LF; no HTML in any value