Add Serply as an optional keyed backend for web and Google Scholar search #321

Context

Stage 4's web augmentation (researchclaw/web/) has two search paths today:

• WebSearchClient: Tavily when TAVILY_API_KEY is set, otherwise a DuckDuckGo HTML scrape.
• GoogleScholarClient: the scholarly scraper (optional web extra). Google blocks it fairly quickly on unattended runs, and use_proxy only partly helps.

There is no keyed option for Google web or Google Scholar results, and there is no way to get Scholar results at all without installing scholarly.

Proposal

Add Serply as an optional keyed backend behind the existing clients, following the same pattern Tavily uses (key from config or env, everything unchanged when the key is absent):

• WebSearchClient: try Serply's /v1/search/ after Tavily and before the DuckDuckGo fallback.
• GoogleScholarClient: query Serply's /v1/scholar/ first when a key is set, fall back to scholarly if that call fails. This also lets the Scholar client run without scholarly installed.
• WebSearchConfig: serply_api_key / serply_api_key_env (default SERPLY_API_KEY), passed through _literature.py to WebSearchAgent like tavily_api_key_env.
• Plain urllib, no new dependency. Tests with mocked responses for both clients plus the config parsing.

Disclosure: I work with Serply. Happy to adjust the shape (for example, a separate scholar_backend setting) if you would rather not have it inside the existing clients.

I have a branch ready and will open the PR referencing this issue.