feat: parse w:wordWrap on paragraph properties

[developer0hye]

What

Parse w:pPr/w:wordWrap into ParagraphProperty::word_wrap.

Why

w:wordWrap is Word's switch between word-level and character-level line breaking for East Asian text: w:val="1" keeps a Hangul eojeol whole, w:val="0" allows a break at any syllable. The reader dropped the element, so a consumer had no way to distinguish a paragraph that explicitly asks for character-level breaking from one that says nothing and leaves the choice to the style chain.

Needed by developer0hye/office2pdf#730, where the value is measured to be the exact discriminator — including overriding the style chain, so it cannot be inferred from pStyle.

Note on the false case

The flags around it (keepNext, keepLines, pageBreakBefore, widowControl) record only the true case:

if read_bool(&attributes) {
    p.widow_control = Some(true);
}

That is fine where 0 and absent mean the same thing. Here they do not — 0 is the setting that carries the instruction — so this arm records both:

p.word_wrap = Some(read_bool(&attributes));

read_bool already returns true for a bare <w:wordWrap/>, which matches what OOXML means by the element with no w:val.

Tests

Two, in the existing reader::paragraph_property test module and following the style of the shading test there:

• w:val="0" survives as Some(false)
• a bare <w:wordWrap/> and w:val="1" both give Some(true), and an absent element gives None, so the three states stay distinct

cargo test -p docx-rs --lib passes 275. The 20 failures in tests/reader.rs are present on fix/parse-tolerance before this change as well — identical count with and without it — so they are untouched by this PR.

Scope follows the shading PR (#2): reader and struct only, no writer.

[developer0hye]

CI status, for the record before merging.

• Clippy, DCO, build: pass.
• test: the 308 lib tests pass, including the two added here. The job is red on the 20 tests/reader.rs integration tests, which fail identically on fix/parse-tolerance without this change — I measured both locally (0 passed; 20 failed either way) and the branch's own CI has been failing since 2026-07-21, before this work.
• build-wasm: fails on tsconfig.node.json(4,25): error TS5108: Option 'moduleResolution=node10' has been removed, a TypeScript toolchain drift unrelated to Rust changes.

Neither red job is caused by this PR, and neither is in scope for it. Merging on that basis.