chore(gallery): add Humanlike Chat Q5/Q6 variants

[localai-org-maint-bot]

Description

Add Q5_K_M and Q6_K llama.cpp variants for Qwen3.8-27B Humanlike Chat and document how to select them. They fill the gap between the existing Q4_K_M and Q8_0 builds.

Both variants pin revision ba6d29fb2505241d7dae88df4fda9038999c3ca9, matching the existing step-576 adapter merge at 0.7 strength. The current upstream card describes a newer release; using the existing revision keeps the variant group consistent. Runtime settings match the existing Q8 entry, including text-only use and disabled thinking.

Notes for Reviewers

• Verified both SHA256 values and byte sizes against independent HF LFS metadata and x-linked-etag / x-linked-size headers at the pinned revision.
• Parsed all 2,085 gallery entries: unique names, valid variant targets, no nested variants or multiply claimed targets.
• Confirmed all 2,082 unrelated entries remain unchanged; the parent only gains two variant references.
• git diff --check passes.
• go test ./core/gallery -run Test -count=1 cannot start: generated pkg/grpc/proto files are absent. No inference or full model downloads were run.
• Publisher source: https://huggingface.co/LessThanThreeAI/Qwen3.8-27B-Humanlike-Chat-GGUF/blob/ba6d29fb2505241d7dae88df4fda9038999c3ca9/README.md

Signed commits

• Yes, I signed my commits. Human DCO sign-off is required; repository policy prohibits AI sign-off.
• Documentation updated (docs/content/) for user-facing changes, or not applicable