refactor(ml): consolidate 21 training binaries into 2 unified baselines

Replace 20 per-model training examples with:
- train_baseline_rl: DQN + PPO (renamed from train_baseline)
- train_baseline_supervised: TFT, Mamba2, Liquid, TGGN, TLOB, KAN, xLSTM, Diffusion
  via model factory + UnifiedTrainable generic training loop

Update Dockerfile.training (16→7 binaries), train.sh MODEL_BINARY map,
and job-template.yaml default. -12,759 lines.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

crates/ml/Cargo.toml

@@ -205,8 +205,12 @@ path = "examples/evaluate_ppo.rs"
 required-features = ["cuda"]
 
 [[example]]
-name = "train_baseline"
-path = "examples/train_baseline.rs"
+name = "train_baseline_rl"
+path = "examples/train_baseline_rl.rs"
+
+[[example]]
+name = "train_baseline_supervised"
+path = "examples/train_baseline_supervised.rs"
 
 [[bench]]
 name = "microstructure_bench"