chore: keep scratch files out of the repository root

Four files had been sitting in the root for two days: saved web pages and the
one-off script that processed them. Untracked, so nothing was ever committed —
but the root of an engine is the first place a newcomer reads, and anything
there is taken for structure until proven otherwise.

The patterns are anchored to the root on purpose. A crate that genuinely serves
HTML must not have its assets silently ignored, so `/*.html` and not `*.html`.

Claude-Session: https://claude.ai/code/session_01PzJTTrKcGrKtCZsGC3Z7S1

Author: ApiliumDevTeam

.gitignore

@@ -137,4 +137,17 @@ contexto/*
 # Local neural-embedder test models (never commit — ~470MB fp32 e5-small, fetched on demand)
 crates/ineru/test-models/
 # Local ONNX Runtime dylib for running the gated neural tests (set ORT_DYLIB_PATH to it)
-.tmp-ort/
+.tmp-ort/
+# Scratch files left in the repository root.
+#
+# Saved web pages and one-off scripts accumulate here during investigation and
+# then sit for days looking like they belong. They do not: the root of an
+# engine is the first place a newcomer reads, and anything there is taken for
+# structure. Keeping a saved copy of a page is also its own trap — the next
+# person checks the copy instead of the source it came from.
+#
+# Deliberately anchored to the root: a crate that genuinely serves HTML must
+# not have its assets silently ignored.
+/*.html
+/*.py
+/scratch/