Exercise 3: A Comment Containing a Literal '<' — Possible Solution
====================================================================

THE TEST
------------------------------
    tokenize("<!-- a < b --><p>after</p>")

RESULT
------------------------------
    [comment(' a < b '), opentag('p', {}), text('after'), closetag('p')]

The comment's own text is exactly ' a < b ' -- the literal '<'
character preserved as ordinary comment content -- and the following
<p> element tokenizes completely normally afterward.

WHICH PART OF tokenize HANDLES THIS
------------------------------
    if html.startswith('<!--', i):
        end = html.find('-->', i + 4)
        comment_text = html[i+4:end]
        i = end + 3
        tokens.append(Token('comment', text=comment_text))
        continue

The key line is `end = html.find('-->', i + 4)`. Once the main loop
recognizes the four-character sequence "<!--" at the current
position, it hands control ENTIRELY to a single, dedicated search:
"find the next occurrence of the literal three-character string
'-->' anywhere after this point." That search has no separate
awareness of '<' or '>' as individually meaningful characters at
all -- it's looking for one exact substring, full stop. The stray
'<' in " a < b " is just one character among many sitting inside the
range being copied into comment_text; it's never inspected on its
own by anything that treats '<' as "the start of a tag," because
that logic lives entirely in the OUTER while loop (`if html[i] !=
'<': i += 1; continue`), which this comment-handling branch never
returns to until after `i` has already been advanced past the full
"-->" delimiter.

WHY THIS WORKS AS AN ANSWER
------------------------------
This is the same underlying design the chapter's own '>' example
demonstrates, just triggered by the other character a naive scanner
might reasonably assume is special. Both cases confirm the same
thing: once inside a comment, the tokenizer's own notion of "what
counts as a delimiter" narrows to exactly one specific string
('-->'), and every other character -- including the two characters
that are normally the most structurally significant in the entire
format -- is just data until that string is found.