fix(fetchers): cap direct llms bodies

[chaliy]

What

Make the direct llms.txt fetch path honor FetchKit's configured body-size cap and truncation semantics.

Closes #95.

Why

The direct llms.txt branch in DocsSiteFetcher read the full response body with .text().await, so it ignored max_body_size, returned oversized payloads intact, and skipped the usual truncation signal.

How

• reuse the default fetcher's bounded body reader and shared truncation marker for direct llms.txt responses
• apply the configured max_body_size, defaulting to the shared 10 MiB cap when unset
• add a regression proving direct llms.txt requests truncate and report truncated: true when the cap is exceeded

Risk

• Low
• direct llms.txt responses now surface size and truncation metadata consistent with the default fetcher; the main regression risk is around callers that assumed uncapped bodies on this specialized path

Checklist

• Unit tests are passed
• Smoke tests are passed
• Documentation is updated
• Specs are up to date and not in conflict