docs(metadata): align launcher with governed execution

[abrichr]

What changed

• align the launcher distribution summary with the governed-execution product
• replace generic ML/VLM/agent keywords with workflow, verification, desktop, RDP, and Citrix discovery terms
• remove research and generic AI classifiers from the production launcher metadata

Why

The current README already presents a coherent flagship entry point and links the canonical engine correctly, but the published package metadata still frames the launcher as ML/VLM/agent research. This keeps the next release's PyPI entry aligned with the actual OpenAdapt product without changing runtime behavior or triggering a release by itself.

Validation

• 44 focused package, CLI, metadata, and release-lock tests passed; 5 platform-dependent checks skipped
• built wheel and sdist passed scripts/verify_release_artifacts.py
• Ruff and git diff --check passed
• inspected generated wheel metadata directly
• verified all public README destinations and the product screenshot return HTTP 200