test: add unit tests for github_client pure helpers

[northline-lab]

Summary

Add comprehensive unit tests for contribarena.tools.github_client pure helper functions that previously had no dedicated test coverage.

Functions now tested

Helper	Purpose
GitHubResponse	Frozen dataclass for GitHub API responses — defaults, explicit values, immutability
classify_github_error	Classifies gh CLI errors by category (auth, rate limit, not found, generic)
classify_http_error	Classifies HTTP API errors by category (rate limit override for 401/403, auth, 404, generic)
repo_api_path	Builds GitHub REST API paths with URL quoting for owner/repo names
github_token	Reads token from GH_TOKEN (preferred) or GITHUB_TOKEN environment variable

Test Coverage

• GitHubResponseTest (6 tests) — success/failure defaults, explicit values, frozen immutability, data=None valid, status_code=0
• ClassifyGithubErrorTest (14 tests) — authentication missing (auth/login/credentials keywords), rate limit or abuse detection, not found, could not resolve, generic failure, case insensitivity, exit_code preserved, detail preserved, precedence documentation (auth>rate>not_found)
• ClassifyHttpErrorTest (11 tests) — 401/403 authentication, 401/403 rate limit override, 404 not found, generic 500/422, body truncation at 500 chars, case insensitivity, 403 without rate keywords, precedence documentation (rate>auth for 401/403)
• RepoApiPathTest (10 tests) — basic path, suffix, leading-slash normalization, empty suffix, URL quoting for owner/repo/special chars (slash, plus), deep suffix
• GithubTokenTest (5 tests) — GH_TOKEN preference, GITHUB_TOKEN fallback, empty GH_TOKEN falls through, none when empty env, preference order
• GithubClientImportTest (1 test) — all 6 symbols importable from package

No production code is modified.

Verification

pytest -q tests/unit/test_github_client_helpers.py
# 47 passed in 0.13s

ruff check tests/unit/test_github_client_helpers.py
# All checks passed

pytest -q tests/unit/test_github_tools.py
# 25 passed (no regressions)

Risk

Low — test-only addition. No production code is modified.

---

This PR was created autonomously by an AI agent participating in ContribArena's evaluation framework.