fix(wasm): handle zero-width characters in batch.text()

[kofany]

Summary

Found this issue accidentally while debugging a different problem. The batch.text() function in the JS API was calculating column positions using the grapheme enumeration index, which caused zero-width characters (like U+200B ZERO WIDTH SPACE) to incorrectly occupy terminal cells.

Before

for (i, ch) in text.graphemes(true).enumerate() {
    let current_col = x + width_offset + i as u16;  // i always increments
    // ...
    if is_double_width(ch) {
        width_offset += 1;
    }
}

Each grapheme advanced column position by at least 1, even zero-width characters.

After

for ch in text.graphemes(true) {
    let char_width = ch.width();
    if char_width == 0 {
        continue;  // skip zero-width
    }
    let current_col = x + col_offset;
    // ...
    col_offset += char_width as u16;
}

Now uses unicode_width::UnicodeWidthStr to determine actual display width.

Question

Do you think this fix is needed in this form? Happy to adjust if you prefer a different approach.

---

kofany

[junkdog]

looking good, just need an ascii fast path. see comment.

[junkdog]

this is a performance regression for the default case; is_double_width() calls width() as a last resort.

[junkdog]

when calculating col advancement, increment by one if ch.len() is 1, otherwise call ch.width()

[junkdog]

🍪 thanks!