fix(agent): include SystemParts in token estimation and add reasoning guards

[badgerbees]

📝 Description

This PR fixes a deficiency in the token estimation logic for context pruning. Currently, the agent ignores SystemParts (structured system blocks used for cache-aware adapters and large instructions), leading to underestimated token counts and potential context window overflows (400 errors). It also streamlines the reasoning content path for better robustness.

🗣️ Type of Change

• 🐞 Bug fix (non-breaking change which fixes an issue)
• ✨ New feature (non-breaking change which adds functionality)
• 📖 Documentation update
• ⚡ Code refactoring (no functional changes, no api changes)

🤖 AI Code Generation

• 🤖 Fully AI-generated (100% AI, 0% Human)
• 🛠️ Mostly AI-generated (AI draft, Human verified/modified)
• 👨‍💻 Mostly Human-written (Human lead, AI assisted or none)

🔗 Related Issue

N/A

📚 Technical Context (Skip for Docs)

• Reference URL: N/A
• Reasoning: In estimateMessageTokens, msg.SystemParts was being ignored. Large instructions or RAG indices kept in system messages would stay hidden from the budget checker, causing overflows on the next provider call.

🧪 Test Environment

• Hardware: PC
• OS: Windows 11
• Model/Provider: Anthropic Sonnet 3.7
• Channels: CLI / Any

📸 Evidence (Optional)

Click to view Logs/Screenshots

=== RUN   TestEstimateMessageTokens_SystemParts
--- PASS: TestEstimateMessageTokens_SystemParts (0.00s)
PASS
ok      github.com/sipeed/picoclaw/pkg/agent    0.340s

☑️ Checklist

• My code/docs follow the style of this project.
• I have performed a self-review of my own changes.
• I have updated the documentation accordingly..