[Optimization]: Implement dynamic Context-Window Pruning and Token Truncation Middleware for Semantic Workspaces

## 🚀 Architectural Optimization Request

### Is your feature request related to a problem?
Currently, when the platform runs workspace search routines or parses extensive meeting files for semantic embedding mapping, large structural data aggregates are compiled directly into the LLM context wrapper payload. If a user queries a broad workspace containing sprawling multi-file histories, the raw data injection risks overflowing the downstream provider's maximum token context window capability.

Even if it doesn't overflow, passing repetitive, low-relevance structural data sequences creates context dilution, drives down summary accuracy, and spikes external API compute/token billing overhead.

### Proposed Solution
Introduce an abstract context preprocessing utility layer. This middleware will dynamically rank retrieved textual chunks using a lightweight localized score filter (like BM25 or a localized Cosine Similarity threshold) and prune non-essential token sequences before compiling the final payload.

### Proposed Implementation Details
1. **Dynamic Content Scoring Layer:** Implement a preprocessing step that reads structural context fragments and matches them against user intent vectors, assigning relevance scores to individual paragraphs.
2. **Hard Token Limit Safeguard:** Build a dynamic sliding window truncation loop. If aggregate chunks exceed a configured max boundary parameter (e.g., `max_workspace_tokens: 4000`), it gracefully drops lowest-scoring sections instead of letting the entire downstream API call crash with a payload error.
3. **Debug Logger Tracking:** Integrate a system tracking matrix that logs the raw compressed efficiency parameters (e.g., `Tokens Reduced: X% | Latency Optimizations: Yms`).

### Additional Context
Implementing this optimization directly protects the application layer against massive enterprise-scale data inputs, drops average network payload weight, and saves operational API resource budgets.