Fix parsing of atypical page numbers and serial numbers

[YDX-2147483647]

• Make Numeric::from_str refuse to parse irregular page numbers like 11E720, so that they can be interpreted as non-numeric strings.
• Let Numeric store leading zeros (e.g., 011) and whitespaces between affixes and numbers (e.g., GB/T 7714) in prefix.
• Add support for <if is-numeric="page">. Previously, it's always false.

Resolves #170
Resolves #440 (which is a part of #312)
Also relates to #327 because number_{LeadingZeros,MixedText} now pass.

---

Notes for Chinese:
This PR eliminates num, code_space, and han_space errors. According to the survey Hayagriva对GB/T 7714—2015的支持情况, they account for 41% of the errors in GB/T 7714—2015.

[YDX-2147483647]

CSL spec says nothing about 1.36, and citeproc-js (citext) thinks it's non-numeric.

Test code

#import "@local/citext:0.3.0": *

#let source = ```bib
@book{a,
  title = {A},
  pages =  {1a},
}
@book{b,
  title = {B},
  pages =  {1.36},
}
```.text

#let csl = ```xml
<?xml version="1.0" encoding="utf-8"?>
<style xmlns="http://purl.org/net/xbiblio/csl" class="note" version="1.0" name-as-sort-order="all" sort-separator=" " demote-non-dropping-particle="never" initialize-with=" " initialize-with-hyphen="false" page-range-format="expanded" default-locale="zh-CN">
  <info>
    <id />
    <title />
    <updated>2009-08-10T04:49:00+09:00</updated>
  </info>
  <citation>
    <layout/>
  </citation>
  <bibliography>
    <layout delimiter="; ">
      <group delimiter=" ">
        <text variable="title"/>
        <text variable="page"/>
        <choose>
          <if is-numeric="page">
            <text value="(is-numeric)"/>
          </if>
          <else>
            <text value="(non-numeric)"/>
          </else>
        </choose>
      </group>
    </layout>
  </bibliography>
</style>
```.text

#let bib = init-citation(source, mode: "stable", csl: csl)
#show: show-extcite.with(bib: bib, gen-id: true)

- @a
- @b

#show bibliography: none
#bibliography(bytes(source))

#extbib(bib)