fix: drop CrossrefProcessor that corrupts URL fragments in docstrings (#83)

pydoc-markdown's `CrossrefProcessor` is run with `resolver=None` in
`generate_ast.py`, so it can never resolve a `#ref` into a real link —
it only falls back to wrapping the ref in backticks. Its regex
`\B#([\w\d._]+)` also matches any `#word` preceded by a non-word char,
so URL fragments like
`.../dockerfile_best-practices/#leverage-build-cache` (the `/#`
sequence) get corrupted into broken links such as
``...best-practices/`leverage`-build-cache``.

The Apify docstrings don't use `#ref` cross-references, so this drops
`CrossrefProcessor` from the processor list — it brought no benefit
(never produced links) and only corrupted fragments. This also matches
the griffe-based extractor, which doesn't run it either.

Verified by running the patched `generate_ast.py` against
apify-client-python: the generated AST now contains
`#leverage-build-cache` intact.

---------

Co-authored-by: Jindřich Bär <jindrichbar@gmail.com>

Author: vdusek

python-scripts/docspec-gen/generate_ast.py

@@ -7,7 +7,6 @@
 from pydoc_markdown.interfaces import Context
 from pydoc_markdown.contrib.loaders.python import PythonLoader
 from pydoc_markdown.contrib.processors.filter import FilterProcessor
-from pydoc_markdown.contrib.processors.crossref import CrossrefProcessor
 from google_docstring_processor import ApifyGoogleProcessor
 from docspec import dump_module
 
@@ -43,15 +42,13 @@ def main():
         documented_only=False,
         skip_empty_modules=False,
     )
-    crossref = CrossrefProcessor()
     google = ApifyGoogleProcessor()
 
     loader.init(context)
     filter.init(context)
     google.init(context)
-    crossref.init(context)
 
-    processors = [filter, google, crossref]
+    processors = [filter, google]
 
     dump = []