Error when `llms.txt` contains level-2 sections and no links

If the `llms.txt` document contains a level-2 section, `parse_link()` throws if there aren't any links in the document. Here's an example using a variation of a test case:

```bash
cat << 'EOF' > llms.txt
# No Links Title

## Description

> No description

Some details without links
EOF

llms_txt2ctx llms.txt
```

```bash
Traceback (most recent call last):
File "/Users/garrick/work/external/AnswerDotAI/llms-txt/.venv/bin/llms_txt2ctx", line 8, in
sys.exit(llms_txt2ctx())
~~~~~~~~~~~~^^
File "/Users/garrick/work/external/AnswerDotAI/llms-txt/.venv/lib/python3.13/site-packages/fastcore/script.py", line 121, in _f
return tfunc(**merge(args, args_from_prog(func, xtra)))
File "/Users/garrick/work/external/AnswerDotAI/llms-txt/llms_txt/core.py", line 125, in llms_txt2ctx
ctx = create_ctx(Path(fname).read_text(), optional=optional, n_workers=n_workers)
File "/Users/garrick/work/external/AnswerDotAI/llms-txt/llms_txt/core.py", line 112, in create_ctx
d = parse_llms_file(txt)
File "/Users/garrick/work/external/AnswerDotAI/llms-txt/llms_txt/core.py", line 57, in parse_llms_file
start,sects = _parse_llms(txt)
~~~~~~~~~~~^^^^^
File "/Users/garrick/work/external/AnswerDotAI/llms-txt/llms_txt/core.py", line 51, in _parse_llms
sects = {k: _parse_links(v) for k,v in d.items()}
~~~~~~~~~~~~^^^
File "/Users/garrick/work/external/AnswerDotAI/llms-txt/llms_txt/core.py", line 45, in _parse_links
return [parse_link(l) for l in re.split(r'\n+', links.strip()) if l.strip()]
~~~~~~~~~~^^^
File "/Users/garrick/work/external/AnswerDotAI/llms-txt/llms_txt/core.py", line 41, in parse_link
return re.search(pat, txt).groupdict()
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'groupdict'
```

Start with llms_txt/core.py, especially parse_link() and _parse_links(), then run the llms_txt2ctx example from the issue with an llms.txt file containing a level-2 section but no links. Done means that input is parsed without the AttributeError and the command completes successfully.

Written by the indexing model from the issue text.