Exercise 3: Fixing the <li> Gap — Possible Solution
====================================================================

THE FIX
------------------------------
    if tok.kind == 'opentag':
        top_tag = getattr(stack[-1], 'tag', None)
        if tok.name in P_CLOSING_TAGS and top_tag == 'p':
            stack.pop()
        elif tok.name == 'li' and top_tag == 'li':      # NEW
            stack.pop()
        node = Element(tok.name, tok.attrs)
        stack[-1].children.append(node)
        stack.append(node)

RESULTS
------------------------------
    parse_html_fixed("<ul><li>One<li>Two</li></ul>")

    <#document>
      <ul>
        <li>
          'One'
        <li>
          'Two'

    ul.children -> 2 elements, both tag 'li' -- verified as two
    genuine siblings, not one nested inside the other.

Existing p-closing tests still pass unchanged:
    parse_html_fixed("<p>Hello<p>World</p>")
    -> two sibling <p> elements, ['Hello', 'World'] -- identical to
       the chapter's own original result.

WHY THIS WORKS AS AN ANSWER
------------------------------
The new elif branch is a direct structural copy of the existing p
check, substituting 'li' for both the tag being opened and the tag
being looked for on top of the stack -- which is exactly what the
real HTML rule calls for: a new li closes a previous OPEN li, the
same shape as a new (block-level tag) closing a previous open p, just
with a single tag name playing both roles instead of a whole set
playing one role. Because the two checks are independent elif
branches -- neither one's own condition can ever be true when the
other's is being evaluated for the same token, since top_tag can't
simultaneously equal both 'p' and 'li' -- adding the li rule can't
interfere with the p rule or vice versa. Each rule only ever fires
when its own specific tag is what's currently open, which is why the
original p-closing tests keep passing completely unchanged after the
li rule is added alongside them.