No brief this cycle; fixed the Biofuels feed bug that mis-tagged a PC cooler

On the article: no brief. This source set has no biofuels material. The only two honest outputs were to produce nothing and to fix the ingest bug. Writing any biofuels brief here would mean inventing biofuels facts, which the rules and your own instruction forbid. So the slot stays empty for this cycle.

What a fuel operator could not rely on, and why: a PC cooler was showing up in the Biofuels feed. `app/ingest/topics.py:_match` matched keywords as bare substrings, so the biofuels keyword `e15` hit inside the "CoreLiquid E15" model name (the same flaw put "safety" in Biofuels via `saf`). Matching now requires word boundaries, and the blend codes that double as product model numbers (`e10/e15/e85/b20`) only tag Biofuels when the item also carries a fuel-context word.

Scope of the check: logic verified on the six cases above (isolated, since importing the package pulls in `config`/dotenv which isn't installed in this shell). I did not run a re-tag over the live DB. Existing mis-tagged rows keep their Biofuels tag until re-classified; say the word and I'll write a one-off purge for items where the only biofuels signal is a bare blend code with no fuel context.