split-pdf: extract splitter into scripts/ and switch backend to pypdf

[nsmiller2501]

Summary

Follow-on to #5. Two changes to /split-pdf:

1. Extract the inline PDF-splitting logic into a standalone Python script at .claude/skills/split-pdf/scripts/split.py.
2. Swap the PDF backend from PyPDF2 to pypdf (drop-in API).

What changed

• New .claude/skills/split-pdf/scripts/split.py: takes the previously inline PdfReader / PdfWriter chunking logic and packages it as a reusable CLI script.
• SKILL.md shrinks: the deterministic chunking steps are replaced with a single script invocation; only the protocol prose remains.
• All PyPDF2 references (script import, SKILL.md install hint, README copy in both .claude/skills/ and skills/) updated to pypdf.

Why

• Script extraction: keeps the deterministic chunking logic out of the LLM context, makes it independently testable, and matches the pattern used by other skills (e.g. /bibcheck/scripts/split_bib.py).
• PyPDF2 → pypdf: PyPDF2 is unmaintained; pypdf is its maintained successor and exposes a drop-in-compatible PdfReader / PdfWriter API. The migration is a one-line import change with no behavior difference for fixed-size chunking.

Testing

• Ran python3 .claude/skills/split-pdf/scripts/split.py path/to/sample.pdf and confirmed the resulting chunk layout matches the prior inline behavior.
• Verified pip install pypdf works on Python 3.10+ and the script imports cleanly.
• No behavioral change in the split output expected.