feat(mcp): add read-only Feishu and Confluence connectors

[juemimgcd]

Summary

Add standalone read-only stdio MCP servers so agents can browse, search, and read pages from a host-configured Feishu Wiki or Confluence Cloud space.

• Feishu: list child nodes, search Wiki, and read docx text excerpts with source URLs and content hashes.
• Confluence Cloud: list pages or direct children, search with space-scoped CQL, and read static page text with source URLs, versions, and content hashes. Support personal API tokens through the site or scoped-token gateway.
• Keep credentials and space selection under host control, validate returned scope, support pagination, and report access/API failures as tool errors.
• Add setup and client configuration documentation. Reuse the existing MCP dependency file.

Validation

• Ruff checks passed; Python compilation passed during implementation.
• Existing dependency configuration tests: 13 passed, including after rebasing onto current master.
• Actual stdio initialization and tools/list succeeded for both servers; each exposes three read-only tools with structured output schemas.
• git diff --check passed.
• No test files were added or changed.

Limitations

Live Feishu and Confluence authorization/search/content retrieval have not been exercised because real credentials and target spaces were not configured. Feishu body reading supports docx only. Confluence supports Cloud only and extracts static text without executing macros or reading attachments/images. Profile-guided retrieval and synchronization into Reminder are outside this change.