fix: training profile must not override GPU profile batch_size

dqn-production.toml batch_size=8192 was overriding GPU profile batch_size=64
on RTX 3050 via apply_to(), causing OOM/hang. GPU profile is the sole
authority for batch_size — training profile controls learning params only.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

config/gpu/rtx3050.toml

@@ -1,6 +1,6 @@
 # RTX 3050 Ti (4GB VRAM) -- minimal footprint
 [training]
-batch_size = 128
+batch_size = 64
 num_atoms = 11
 buffer_size = 0  # 0 = auto from VRAM
 hidden_dim_base = 256

crates/ml/src/training_profile.rs

@@ -670,9 +670,8 @@ impl DqnTrainingProfile {
             if let Some(v) = t.epochs {
                 hp.epochs = v;
             }
-            if let Some(v) = t.batch_size {
-                hp.batch_size = v;
-            }
+            // batch_size: NOT applied from training profile.
+            // GPU profile (config/gpu/*.toml) is the sole authority for batch_size.
             if let Some(v) = t.learning_rate {
                 hp.learning_rate = v;
             }