Pagelist parsing breaks on non-numeric page numbers and does not use hrefs.

[martinq]

When parsed using either ncxItem or item, the page number text is always run through the parseInt function. This fails on non-numeric page numbers like Roman numerals in prologues, e.g.:

<nav xmlns:epub="http://www.idpf.org/2007/ops" epub:type="page-list" hidden="hidden">
    <ol>
        <li><a href="Titlepage.xhtml#page_i">i</a></li>
        <li><a href="Dedication.xhtml#page_iii">iii</a></li>
        <li><a href="Foreword.xhtml#page_v">v</a></li>
        <li><a href="Foreword.xhtml#page_vi">vi</a></li>

In addition, the only page information that is retained from either type of item is the page number itself and any associated CFI. The href property is not used. As a result, page list navigation will only function if pages are expressed in CFI.