⚡ v4.3: Performance optimizations — ONNX INT8, BGE embedder, batched classification, thread control (#4)

Fix: DeBERTa-v3 fp16 crash on T4 — use fp32, batch=2, grad_accum=16

Update notebook for T4 GPU: batch=4, grad_accum=8, eval_batch=8

Add ClauseGuard v4 Colab notebook: DeBERTa-v3-large 2-stage training (LEDGAR→CUAD) with ASL

Update ML requirements for DeBERTa-v3-large training

Add ClauseGuard v4 training script (DeBERTa-v3-large + LEDGAR + CUAD + ASL)

v2: Fix all versions + minimal UI redesign (no gradients, clean layout)

Add Colab training notebook

Fix: cast labels to float32 for BCEWithLogitsLoss compatibility

🛡️ ClauseGuard v1.0 — Full codebase: Extension + Website + API + ML