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.

.gitignore

@@ -0,0 +1,23 @@
+.venv/
+swanlog/
+.pytest_cache/
+__pycache__/
+*.py[cod]
+ckpts/
+data/spm_4k.*
+data/pretrain/*
+!data/pretrain/.gitkeep
+data/sft/*
+!data/sft/.gitkeep
+!data/sft/toy.jsonl
+data/eval/flores*
+
+# LaTeX build output
+*.aux
+*.log
+*.out
+*.toc
+*.fls
+*.fdb_latexmk
+*.xdv
+wiki jsonl