when computing <ol><li> numbering, ignore non-<li> previous siblings

[chrispy-snps]

Currently, numbering for <li> items inside <ol> count all previous siblings when computing the number:

import markdownify

html = """
<ol>
  <!--comment-->
  <li>a</li>
  <div/>
  <li>b</li>
</ol>
"""

print(markdownify.MarkdownConverter().convert(html))
# 2. a
# 4. b

This pull request updates the code to consider only preceding <li> elements:

print(markdownify.MarkdownConverter().convert(html))
# 1. a
# 2. b

A unit test is added to check this case.