feat: add data flow annotations to connections

[trmquang93]

Summary

• Connections now carry typed data items (dataFlow: [{ name, type, description }]) describing what information passes between screens
• New DataFlowEditor component integrated into both ConnectionEditModal and HotspotModal (navigate, modal, API success/error, and conditional modes)
• Teal pill badge on canvas connections that have data annotations
• Generated instructions now include "Input Parameters" tables per screen in screens.md, a "Data" column in navigation.md, and annotated wire items in tasks.md
• MCP server create_connection and update_connection tools accept data_flow parameter
• FILE_VERSION bumped from 10 to 11 with backward-compatible migration

Why

This is the single biggest gap in instruction generation. The flow captures where users navigate but not what data they carry. Without this, AI agents generate screens that don't receive the data they need, leading to broken navigation or redundant API calls.

Test plan

• Create a connection between two screens, open the edit modal, add data items (name, type, description) -- verify they persist after save and reopen
• Verify teal badge appears on connections with data, shows param name (1 item) or "N params" (multiple)
• Open HotspotModal for navigate/modal action -- verify DataFlowEditor appears after target screen
• Open HotspotModal for API action -- verify DataFlowEditor in On Success and On Error sections
• Export instructions -- verify navigation.md has "Data" column and screens.md has "Input Parameters" tables
• Open an older .drawd file (v10) -- verify it loads without errors (dataFlow backfilled to [])
• MCP server: create_connection with data_flow param, verify in list_connections
• npm run lint passes
• npm test passes (271/271, excluding pre-existing @grida/refig issue)