Initial import: XAUUSD news sentiment models, evals, and data tooling

TF-IDF baseline (news_sentiment.py, models/xauusd_sentiment.joblib) and the
two-head DeBERTa-v3-small transformer (transformer/, news_transformer.py)
with its Google News crawler and DeepSeek labeller (data_collect/), the
golden-set and archive evals (eval/), and the 2026-09-22 and 2026-09-27
results and progress notes.

Transformer weights (model.pt, model.onnx) and the crawl database are not
tracked; rebuild with transformer/train.py and transformer/export_onnx.py.
The DeepSeek label cache in data/ is tracked.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

.gitignore

@@ -0,0 +1,19 @@
+# environments
+.venv/
+.venv-train/
+__pycache__/
+*.py[cod]
+
+# sqlite side files
+*.db-shm
+*.db-wal
+
+# crawl database: 50+ MB and rewritten by every crawl. The paid DeepSeek
+# labels are also in data/history_deepseek_cache.json, which is tracked.
+data/history.db
+
+# transformer weights: 566 MB torch, 271 MB ONNX. meta.json and
+# tokenizer.json are tracked; rebuild with transformer/train.py and
+# transformer/export_onnx.py.
+models/xau_transformer/**/model.pt
+models/xau_transformer/*.onnx