Follow pagination when listing GitHub names

[bIackr0se]

GithubFileSystem.tags, branches and repos() read only the first GitHub API page. When results span several pages, valid names are missing.

Follow the response's rel="next" link until no next page remains. Each request keeps the existing timeout and instance authentication. A later-page HTTP error raises instead of returning a partial list.

Observed against fsspec/filesystem_spec on 2026-10-04:

fs = GithubFileSystem(
    org="fsspec", repo="filesystem_spec",
    sha="778f95612ca60da45ccfb14c22b9ee409d67f1ee",
)
len(fs.tags)
# Before: 30
# After:  103, matching all four API pages

Verification:

• Nine regression cases fail before the fix; all 17 pagination tests pass after it. They cover three-page traversal, exact continuation URLs, authentication, timeouts, empty results, refs and later-page errors.
• Pagination, GitHub path and filesystem specification tests: 247 passed, 67 skipped, 1 expected failure on macOS with Python 3.14.7.
• A wheel installed outside the checkout passes three-page listings and failure paths through a local HTTP server. Against the public API it discovers all 103 tags, then selects 2023.9.2 from a later page and reads its README, matching the raw file byte for byte.
• Repository pre-commit hooks pass.