fix: honor Retry-After without shortening backoff

[0xGhostCasper]

Summary

Companion to the Google AI Mode backpressure changes: configured retryable HTTP errors, including capacity 503 responses, now honor Retry-After delay-seconds or HTTP-dates without shortening the existing exponential backoff. This prevents premature SDK retries under load; it does not fix or establish upstream Google AI Mode success rates.

• Preserve transport backoff, retryable status policy, and maximum retry attempts.
• Reuse the standard-library parser for 429 metadata, preserving RateLimitError for malformed or date headers; 429 remains non-retryable by default.
• Cover the async public AI Mode method and synchronous programs using asyncio.run(), plus GET, GET-with-headers, and POST retry paths.
• Align package, exported version, User-Agent, lockfile, README, and changelog for 0.43.2. Release is pending; this draft PR must not be merged or published yet.

Validation

• 740 non-live tests passed on each of Python 3.10, 3.12, and 3.14.
• 89 focused retry, client, config, and exception tests passed.
• Full-repository Ruff lint and formatting checks passed.
• Mypy passed for all 207 source files.
• Version 0.43.2 wheel and source distribution built successfully.
• No new dependencies, no live scraping calls, and no release performed.

[codecov]

Codecov Report

❌ Patch coverage is 95.23810% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/scrapebadger/_internal/client.py	95.00%	0 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!