Fail fast on GitHub API rate limit instead of hanging for 25+ minutes

[LalatenduMohanty]

Problem

When the GitHub API rate limit is hit, RetryHTTPAdapter._handle_github_rate_limit() in src/fromager/http_retry.py caps wait time at 300s per retry. With the default config (FROMAGER_HTTP_RETRIES=8), this means up to 8 retries × 300s = ~40 minute hang before failing.

In CI, multiple parallel jobs share one IP, quickly exhausting the 60 req/hr unauthenticated limit. The actual reset is ~3600s away, but the 300s cap means the code sleeps, retries, hits the same 403, and repeats.

Proposed Solution

• Add GitHubRateLimitError(RequestException) exception
• In _handle_github_rate_limit(): when wait time exceeds a threshold (e.g. 120s), raise immediately with an actionable message suggesting GITHUB_TOKEN
• Keep existing sleep-and-retry for short waits
• Exclude GitHubRateLimitError from RETRYABLE_EXCEPTIONS so @retry_on_exception on _find_tags won't swallow it

Context

PR #1117 passes GITHUB_TOKEN to CI but tokens aren't available for fork PRs. This ensures fork PR CI fails fast with a clear message instead of hanging.

Ref: https://github.com/python-wheel-build/fromager/actions/runs/25432530885/job/74602126392?pr=1115

Files

• src/fromager/http_retry.py
• tests/test_http_retry.py