Wire up references.bib and verify citations against source PDFs ...

- Add citations/ folder holding the cited papers' PDFs (babapoor2022,
  wu2011, nishi2021), documented in CLAUDE.md so cite keys can be
  filled in/verified directly from the source paper instead of guessed.
- Fill in references.bib entries for all three keys with real
  bibliographic data, and drop a stray "@TODO" in the header comment
  that was making bibtex misparse the file as a malformed entry.
- Switch main.tex from the template's placeholder internal
  bibliography to \bibliography{references}, so citations actually
  resolve (previously rendered as undefined/"?").
- Correct the Introduction's babapoor2022/wu2011 citation claims after
  reading the full papers: babapoor2022 reports separate no-reflow
  (17.1%) and slow-flow (12.9%) rates in a single-center primary-PCI
  cohort, not a general "10-20% of PCI patients" figure; wu2011's
  finding is specific to AMI patients and a composite no-reflow/
  slow-flow (TIMI <=2) endpoint, and only the amount (not presence) of
  attenuated plaque differed significantly.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>