feat: support GitHub tree/<ref> URL for code repository import

[yangxinxin-7]

Background

Previously, only plain repository URLs (https://github.com/user/repo) were recognized as code repositories. URLs containing a specific branch or commit ref (e.g.,
https://github.com/user/repo/tree/main) were routed to HTMLParser and treated as regular web pages, making it impossible to import a specific version of a repository.

Changes

openviking/utils/code_hosting_utils.py

Extended is_git_repo_url to recognize owner/repo/tree/ URLs in addition to plain owner/repo URLs. Only exactly 4-segment paths with tree at position 2 are matched,
avoiding false positives like /blob/, /issues/, or deeper sub-paths.

openviking/parse/parsers/code/code.py

• _parse_ref_from_path: When a tree/ URL is parsed, the ref is now checked against a hex heuristic (_looks_like_sha). If it looks like a commit SHA (7–40 hex
  characters), it is placed in the commit field; otherwise it goes to branch. This ensures non-GitHub hosts use the correct git operation (--branch for branches, fetch +
  checkout for SHAs).
• _github_zip_download: Fixed ZIP URL construction — removed the refs/heads/ prefix so the URL works for branch names, tags, and commit SHAs alike (archive/{ref}.zip).
• parse(): GitHub now always uses the ZIP API regardless of whether a branch or commit is specified, passing branch or commit as the ref.

Usage

branch

client.add_resource("https://github.com/user/repo/tree/main")

commit SHA (short or full)

client.add_resource("https://github.com/user/repo/tree/ab849f44f252e65b2ea3106322c235d8c2f349ad")
client.add_resource("https://github.com/user/repo/tree/ab849f4")