[Logseq] Improve CJK character handling in CLI mode

## Summary `normalize-ident-name-part` in `deps/db/src/logseq/db/frontend/db_ident.cljc` only retains ASCII characters `[0-9a-zA-Z*+!_'?<>=-]`. All CJK (Chinese/Japanese/Korean) characters are silently dropped. In the **Logseq desktop app**, this is masked by the random suffix appended to idents (e.g. `"" + "-a1b2c3d4"` → still unique). But in **CLI and nbb-logseq repeatable mode** (`#?(:org.babashka/nbb true ...)`), no suffix is added — so multiple CJK-only names all normalize to `""`, producing identical `:db/ident` values. ## Reproduction Using `@logseq/cli` (v0.4.3) or `nbb-logseq`: ```clojure ;; Create a graph with multiple CJK class names {:classes {:user.class/项目 {} :user.class/笔记 {} :user.class/标签 {}} :pages-and-blocks []} ``` ```bash logseq import-edn -g my-graph -f data.edn ``` **Result:** ``` Assert failed: All class db-idents must be unique but the following are duplicates: (:user.class/) ``` **Expected:** CJK characters should be preserved in the ident, e.g. `:use