optimize empty-line handling for li and blockquote content

[chrispy-snps]

Fixes #170.

The convert_li() and convert_blockquote() functions are updated to use "substitution functions" to compute how each line should be modified. Inside the functions, different actions are taken for content lines versus empty lines.

For example (spaces replaced with · in output):

from markdownify import markdownify as md

html = """
<blockquote>
  <ul>
    <li>
      <p>1</p>
      <ul>
        <li>
          <p>1-a</p>
          <p>2-b</p>
        </li>
      <ul>
    </li>
    <li>
      <p>2</p>
    </li>
  </ul>
</blockquote>
"""
print(md(html, wrap=True, wrap_width=1e8))
# >·*·1
# >
# >···+·1-a
# >
# >·····2-b
# >·*·2

Unit tests are updated to reflect the changes. In addition, manual tests were performed by using Pandoc to convert the Markdown back to HTML to verify correctness.