fix: decode HTML entities and add ViewBlock renderer in Document renderer

[Copilot]

Two rendering bugs in the Lark Document block renderer: HTML entity characters (&, <, etc.) returned by the Lark API were displayed literally instead of decoded, and image blocks nested inside view blocks silently failed to produce <img> tags.

HTML entity decoding (Text.tsx)

The Lark API returns HTML-encoded content in TextRun.content and link.url. React renders these as plain strings, so & was visible to the user instead of &.

• decodeHTMLEntities uses DOMParser (standard browser API) to decode entities; a single-pass regex over a fixed HTML_ENTITIES map is retained as an SSR (Node.js) fallback
• Applied to both content and link.url in TextRunComponent

// Before
href={link?.url}          // & survives in the DOM attribute
{content + ''}            // "&A" renders as "&A"

// After
href={link?.url ? decodeHTMLEntities(link.url) : undefined}
{decodeHTMLEntities(content + '')}   // "&A" renders as "&A"

Missing <img> tags for view-wrapped images (Layout.tsx, Block.tsx)

Lark document images can be displayed in Card, Preview, or Inline mode, which wraps the image block inside a view block (BlockType.view = 33). Because blockComponentMap had no entry for BlockType.view, ChildrenRenderer silently rendered <noscript> for the view block, and the nested image block was never reached.

Fixed by adding ViewBlockComponent to Layout.tsx and registering it in blockComponentMap:

// Layout.tsx
export const ViewBlockComponent: FC<ViewBlock> = ({ children }) => (
    <ChildrenRenderer>{children}</ChildrenRenderer>
);

// Block.tsx
[BlockType.view]: ViewBlockComponent,

[TechQuery]

有没有针对 HTML 实体的通用解决方案？最好是标准 API。

[Copilot]

Updated in the latest commit to use DOMParser as the standard browser API:

if (typeof DOMParser !== 'undefined')
    return new DOMParser().parseFromString(text, 'text/html').body.textContent ?? text;

The SSR (Node.js) regex fallback is retained since DOMParser is not available there. Using .body.textContent (not .innerHTML) ensures all HTML tags are stripped — the result is always safe plain text.

[TechQuery]

不是网址解码错误，而是普通文本中显示出了 HTML 实体的代码原文。

[Copilot]

Removed safeURL in the latest commit. The href now just decodes entities without extra URL validation:

href={link?.url ? decodeHTMLEntities(link.url) : undefined}

[TechQuery]

图片块中 token 一定存在，我让你排查的是为什么有些图片会完全渲染不出 img 标签。

[Copilot]

The if (image.token) guard has been reverted. After tracing the full rendering pipeline, the actual cause of missing <img> tags is that BlockType.view (33) had no entry in blockComponentMap — so image blocks nested inside a view block (Lark's Card/Preview/Inline display modes) would silently fall through to <noscript> and never be rendered.

Fixed by adding ViewBlockComponent to Layout.tsx and registering it in blockComponentMap:

// Layout.tsx
export const ViewBlockComponent: FC<ViewBlock> = ({ children }) => (
    <ChildrenRenderer>{children}</ChildrenRenderer>
);

// Block.tsx
[BlockType.view]: ViewBlockComponent,