Initial K3 snapshot: 0.5B KDA/MLA/MoE train path

Standalone tree split from LLMRL/projects/kda. Includes Triton dt_bias
backward fix, train_k3 --preset 0.5b, SFT, Docker runtime, and tests.

kda/models/__init__.py

@@ -0,0 +1,7 @@
+"""Configs and the single CausalLM entry."""
+
+from .causal_lm import CausalLM
+from .config import KDAConfig
+from .k3_config import K3Config
+
+__all__ = ["CausalLM", "K3Config", "KDAConfig"]