Simplify WebIntelligence: Remove over-engineered task pipeline and use existing Brain/NewTopics structures

[Copilot]

The WebIntelligence system was overly complex and brittle, making it difficult for users to understand and maintain. This PR dramatically simplifies the system to be "actually usable by a normal human being" as requested.

What was wrong

The original WebIntelligence plugin used a complex multi-stage task pipeline:

1. ObserveAsync → created PlanSearch tasks
2. ExecutePlanAsync → created multiple ExecuteSearch tasks
3. ExecuteSearchAsync → finally performed searches and added to Brain
4. ResearchStore → complex intermediate tracking system
5. Multiple payload types, JSON serialization, correlation keys, and staged messages

This over-engineering made the system brittle and hard to follow.

What this fixes

Simplified Architecture:

• Direct Processing: ObserveAsync now processes unfamiliar topics immediately
• No Intermediate Tasks: Eliminated the complex PlanSearch→ExecuteSearch pipeline
• Uses Existing Structures: Leverages LLMSystem.Bot.Brain.Memories and ChatSession.NewTopics.Unfamiliar_Topics directly
• Built-in Duplicate Prevention: Checks existing Brain memories by topic name before searching

Code Reduction:

• WebIntelligencePlugin.cs: 230 → 177 lines (-23%)
• ResearchStore.cs: Completely removed (-83 lines)
• Total: ~136 lines of complex code eliminated

How it works now

// Simple, direct flow:
1. Check if session has unfamiliar topics
2. For each topic: check if already exists in Brain (prevent duplicates)
3. Skip low-priority topics if budget is low  
4. Perform web search → merge results → add directly to Brain.Memories
5. Done - no complex state tracking needed

The system now automatically processes unfamiliar topics from previous chat sessions, ensures no duplicates in the Brain, and adds knowledge seamlessly in the background. It "just works" without requiring deep understanding of complex task orchestration.

Testing

The simplified logic has been verified to:

• ✅ Prevent duplicate topics using existing Brain.Memories
• ✅ Respect daily search budgets and urgency filtering
• ✅ Process topics directly without intermediate storage
• ✅ Maintain all original functionality with much simpler code

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.