refactor(lsp): reduce client messages to the client to a max of 5 (#25962)

> ## Pull request overview
>
> This PR refactors the language server’s client-notification flow to avoid flooding editors with too many `window/showMessage` notifications by capping the number of messages sent in a batch to 5.
>

Test generated with GPT-5.3-Codex 🤖