fix(nodes): keep the scheme of absolute URLs extracted by ParseNode

[MohammadHijjawi97]

Summary

ParseNode.url_pattern begins with [http[s]?:\/\/]?, which is a character class rather than an optional http(s):// group. The scheme is therefore never part of the match. When parse_urls=True (OmniScraperGraph), every absolute link or image URL loses its scheme and is then joined onto the page URL as if it were relative:

node = ParseNode(input="doc & (url | local_dir)", output=["parsed_doc", "link_urls", "img_urls"],
                 node_config={"chunk_size": 4096, "parse_urls": True, "llm_model": None})
text = "[guide](https://docs.example.org/guide/start) ![logo](https://cdn.example.net/img/logo.png)"
node._extract_urls(text, "https://example.com/blog/post")
# before: (['https://example.com/blog/docs.example.org/guide/start'], ['https://example.com/blog/cdn.example.net/img/logo.png'])
# after:  (['https://docs.example.org/guide/start'], ['https://cdn.example.net/img/logo.png'])

With a local HTML source the same URLs were dropped entirely, because only results starting with http are kept. The broken img_urls are the ones OmniScraperGraph sends to ImageToTextNode.

Changes

• Capture the scheme with (https?:\/\/). The pattern already required ://, so the set of matched URLs is unchanged; they now keep their scheme. Relative URLs are still resolved against the page URL.
• Added tests/nodes/parse_node_test.py covering absolute and relative links and images, a local source, and execute() with parse_urls=True.

Testing

• pytest tests/nodes/parse_node_test.py: 3 failed on pre/beta, 3 passed with the fix.
• pytest tests/test_error_page_detection.py: passes.
• ruff check, black --check and isort --check-only on the changed files, at the versions pinned in .pre-commit-config.yaml.