fix(data): rewrite bundled Skeleton.hwpx id to fit in signed int32

[seonghoony]

Closes #35.

Root cause

src/hwpx/data/Skeleton.hwpx is the seed used by HwpxDocument.new(). Its Contents/section0.xml shipped with

<hp:p id="3121190098" ...>

3121190098 = 0xBA0994D2, with bit 31 set. Every newly-created document inherits this out-of-range id at its first paragraph, and the _allocate_*_id helpers (max + 1) propagate the high value into subsequent ids whenever the skeleton paragraph is the current maximum.

Fix

Replace the lone out-of-range id with 0. The other ids inside the archive (Contents/header.xml, settings.xml, etc.) were already in range, so a single replacement is enough.

-<hp:p id="3121190098" paraPrIDRef="0" styleIDRef="0" pageBreak="0" ...
+<hp:p id="0"          paraPrIDRef="0" styleIDRef="0" pageBreak="0" ...

The repacked archive keeps the original file order and stores mimetype first / uncompressed, matching the OPC convention used by the existing template.

Regression test

tests/test_skeleton_template_ids.py (new):

from importlib.resources import files
import re, zipfile


def test_skeleton_template_ids_fit_in_signed_int32() -> None:
    skeleton = files("hwpx.data") / "Skeleton.hwpx"
    with zipfile.ZipFile(skeleton.open("rb")) as zf:
        for name in zf.namelist():
            if not name.endswith(".xml"):
                continue
            data = zf.read(name).decode("utf-8")
            for m in re.finditer(r'\sid="(-?\d+)"', data):
                value = int(m.group(1))
                assert 0 <= value < 2**31, (
                    f"{name} contains id={value} (0x{value:x}); "
                    "Skeleton.hwpx values must stay in [0, 2^31)"
                )

Without the fix, this test fails on the seed file:

AssertionError: Contents/section0.xml contains id=3121190098 (0xba0994d2);
Skeleton.hwpx values must stay in [0, 2^31)

With the fix:

$ pytest tests/test_skeleton_template_ids.py
tests/test_skeleton_template_ids.py .                                    [100%]
============================== 1 passed in 0.04s ===============================

Wider regression check

$ pytest tests/ -k 'new or skeleton or template or empty'
82 passed, 168 deselected in 0.72s

HwpxDocument.new() and the existing template/regression tests all continue to pass.

Notes

• This PR only touches the bundled binary template plus a small regression test.
• It pairs cleanly with the _paragraph_id / _object_id / _memo_id PR (_paragraph_id / _object_id / _memo_id produce values that overflow signed 32-bit #34) but does not depend on it; either can land first.
• No public API changes.