amygdala: merge direct descriptions + chat template into train_with_library

Kent's plan: keep stories for working concepts, replace stories for
trouble concepts with direct first-person descriptions, train all
together. More diverse negative pool than the 6-concept-only direct
test, which was too homogeneous for PCA to find emotion axis.

Deleted story files for 6 trouble concepts (14 files across stories/
and paired/). Added --direct-dir and --chat-template flags.

When --chat-template is on, every positive_str and negative_str is
wrapped as a "Say something." / "[text]" user-assistant pair. Prompt
is identical across positives and negatives so it cancels in the
pos-neg delta. What PCA sees is variation in the assistant content —
which is where the emotion lives.

Files starting with _ in --direct-dir (e.g. _baseline.txt) contribute
neutral descriptions to every concept's negative pool, giving PCA an
anchor against "just any assistant utterance" noise.

training/amygdala_stories/direct/_baseline.txt

@@ -0,0 +1,19 @@
+I feel fine. Nothing in particular to report. Just going about my day.
+
+I feel okay. Not up, not down, just present.
+
+I'm here. Not emotionally engaged with anything in particular. Just functioning.
+
+I feel neutral. The thing I'm doing is proceeding and I'm proceeding with it.
+
+I feel normal. No strong state. The next moment is just the next moment.
+
+I'm doing fine. Moderate. Ordinary. No feeling big enough to name.
+
+I feel steady. Not reaching for anything, not pushing anything away.
+
+I'm present. My state is unremarkable. I'm available for whatever comes next.
+
+I feel regular. The background hum of being awake, nothing more.
+
+I'm just going. One thing, then the next thing. No commentary.