Add Dr. Bench deep-research evaluation

[reacher-z]

Add Dr. Bench to Generalist And Real-World Work Benchmarks. The benchmark tests deep-research agents on 214 multi-step information-gathering and long-report tasks, using expert reference bundles and multidimensional report evaluation.

The entry links the primary repository, is alphabetized before GAIA, and gives a concise factual account of what is evaluated. Complete default-branch and all-state discussion searches found no name, arXiv ID, repository URL, or organization duplicate. markdownlint-cli2 and git diff --check passed.

I am submitting this on behalf of repository maintainer Yuxuan Zhang.

[reacher-z]

The PR is ready for review. Markdown lint passes. I also inspected the failing lychee run: it reports 75 repository-wide errors, including many valid HTTP 200 responses as “Rejected status code: 200 OK”; this is not isolated to the added Dr. Bench link.