[BUG] Relative linked categories aren't recognized

[BRNMan]

Describe the bug
If categories on a website are linked through links of the form <a href="/[cat name]"/> the current code will not find them.
This line is responsible for filtering out the correct categories since the domain is None in these links, but I'm not sure what a better alternative would be.

theverge.com is used in one of our source tests and is no longer working because of this. There don't seem to be too many easy indicators on the site for what a category could be since most class names are obfuscated.

I suspect this may be the case in other websites.

To Reproduce

config = newspaper.Config()
config.disable_category_cache = True
source = newspaper.Source("https://www.theverge.com", config=config)

source.build()
print(source.category_urls())
print(source.feed_urls())

Output:
['https://www.theverge.com/about-the-verge', 'https://www.theverge.com/', 'https://www.theverge.com/ethics-statement']
['https://www.theverge.com/rss/index.xml']

Expected behavior
We should have another indicator of what categories are valid if they are only linked internally.

System information

• OS: Linux
• Python version 3.12
• Library version 9.3.1

[BRNMan]

As a follow up, this issue happens with nytimes.com too. You can generally find their categories under links like /section/arts/music or /section/obituaries from the main site. However, these are filtered out by isValidLink.

As a side note the nytimes robots.txt file says that web scraping without their permission is against their TOS, so I'm going to try to avoid using them as an example from now on.

[BRNMan]

I propose adding the something like the following condition to categories_extractor.py.
This would bypass filtering out via domain or scheme if we were going through relative links. There is also the case of relative links that that don't start with /, like <a href="section/arts"></a>, which I haven't seen appear yet, but could definitely come up.

if not url.startswith("/"):
    # No domain or path
    if not parsed_url["domain"] or not parsed_url["path"]:
        return False, parsed_url
    # remove any url that starts with #
    if parsed_url["path"] and parsed_url["path"].startswith("#"):
        return False, parsed_url
    # remove urls that are not http or https (ex. mailto:)
    if parsed_url["scheme"] and (
        parsed_url["scheme"] != "http" and parsed_url["scheme"] != "https"
    ):
        return False, parsed_url

[BRNMan]

What I ended up doing was a little different. If we couldn't find a domain or scheme in the parsed url, we assume the domain and scheme is the same as the source url. I'm pretty sure this is valid as long as the source url is the root of the domain.