feat: Context Inspector — sub-agent context visibility, coverage scores, lint warnings

[vivekchand]

Closes #9

What

Implements the Context Inspector feature for debugging context loss in multi-agent AI systems.

Backend ():

• Scans workspace for known context files (SOUL.md, AGENTS.md, MEMORY.md, USER.md, TOOLS.md, etc.)
• Reads session index to build agent tree with depth/parent relationships
• Computes per-agent Coverage Score (0–100) based on: task presence, SOUL.md, MEMORY.md, AGENTS.md, task length ≥50 chars, memory/ files present
• Generates Lint Warnings per agent: context starvation on deep sub-agents, missing USER.md when task references user data, missing SOUL.md/MEMORY.md
• Deduplicates global lint warnings across sessions

Frontend (Context tab):

• Nav tab in both light/dark themes
• 4 stat cards: Total Agents, Avg Coverage %, Lint Warnings, Context Files Found
• Workspace context file checklist with ✅/❌ and KB sizes
• Lint warnings panel with colour-coded severity (🔴 error / 🟡 warn / 🔵 info)
• Agent context tree with depth indentation, coverage score badges, spawn task snippet, and missing file highlights

How

• New blueprint with
• helper with , , sub-functions
• async JS function wired to
• Page HTML added to both theme sections

Tests

5 new tests in — all 81 tests pass (6 skipped unchanged)