Unified language identifiers handling #59

Follow-up to #58.

Two parallel language lists have to stay in sync: extensions/zed/extension.toml and Language in src/document/language.rs. That's pretty fragile to maintain.

Moreover, the language id is not well-defined by LSP, so it's all depends on the implementations(editors). For example: Zed sends language_id = name.to_lowercase() although VS Code sends its own language ids. Currently, enum Language was written mostly against the VS Code id, so it can't be correct for both editors at once.

Some reference:

• https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#textDocumentItem
• https://zed.dev/docs/extensions/languages#multi-language-support
• https://code.visualstudio.com/docs/languages/i