feat: ADR-024 Contrastive CSI Embedding Model — all 7 phases (#52)

Full implementation of Project AETHER — Contrastive CSI Embedding Model.

## Phases Delivered
1. ProjectionHead (64→128→128) + L2 normalization
2. CsiAugmenter (5 physically-motivated augmentations)
3. InfoNCE contrastive loss + SimCLR pretraining
4. FingerprintIndex (4 index types: env, activity, temporal, person)
5. RVF SEG_EMBED (0x0C) + CLI integration
6. Cross-modal alignment (PoseEncoder + InfoNCE)
7. Deep RuVector: MicroLoRA, EWC++, drift detection, hard-negative mining, SEG_LORA

## Stats
- 276 tests passing (191 lib + 51 bin + 16 rvf + 18 vitals)
- 3,342 additions across 8 files
- Zero unsafe/unwrap/panic/todo stubs
- ~55KB INT8 model for ESP32 edge deployment

Also fixes deprecated GitHub Actions (v3→v4) and adds feat/* branch CI triggers.

Closes #50

rust-port/wifi-densepose-rs/crates/wifi-densepose-sensing-server/src/graph_transformer.rs

@@ -486,6 +486,16 @@ impl CsiToPoseTransformer {
     }
     pub fn config(&self) -> &TransformerConfig { &self.config }
 
+    /// Extract body-part feature embeddings without regression heads.
+    /// Returns 17 vectors of dimension d_model (same as forward() but stops
+    /// before xyz_head/conf_head).
+    pub fn embed(&self, csi_features: &[Vec<f32>]) -> Vec<Vec<f32>> {
+        let embedded: Vec<Vec<f32>> = csi_features.iter()
+            .map(|f| self.csi_embed.forward(f)).collect();
+        let attended = self.cross_attn.forward(&self.keypoint_queries, &embedded, &embedded);
+        self.gnn.forward(&attended)
+    }
+
     /// Collect all trainable parameters into a flat vec.
     ///
     /// Layout: csi_embed | keypoint_queries (flat) | cross_attn | gnn | xyz_head | conf_head