feat(gallery): add Maple-Preview GGUF variants

[localai-org-maint-bot]

Description

Add Maple-Preview, DeepGrove's 20B reasoning model with about 1B active parameters, to the gallery. Offer four GGUF variants: TQ1_0 and TQ2_0, each with a Q4_K or F16 output head. The smallest build groups the other three through variants.

Use the publisher's CPU configuration, temperature 1.0, top-p 0.95, and embedded Jinja template. Set an 8K default context within the model's 128K native limit. Add installation and explicit variant examples to the gallery docs.

Notes for Reviewers

• Official GGUF repository, pinned at f5466f918e0c50cdb9d4d47a6f35813509a42a30.
• All four SHA256 values match both Hugging Face LFS metadata and x-linked-etag headers from revision-pinned downloads.
• LocalAI's pinned llama.cpp revision e613ef2c81bae98d59850d061ac29e6e3e88cb00 includes the Maple loader and graph. The inspected GGUF declares general.architecture=maple and the loader's required sliding-window and expert metadata.
• Preserve the embedded template as directed by the publisher's setup guide. No backend changes or GPU claims.
• go test ./core/gallery -count=1 -ginkgo.focus='gallery/index.yaml variant invariants' passes.
• YAML validation confirms all 1,890 existing entries remain unchanged, and the four new entries match the pinned file metadata.
• go test ./core/gallery -count=1: 392 passed, 3 failed, 9 skipped. Unchanged master produces the same three failures fetching external GitHub/gist fixtures with Forbidden responses.
• git diff --check passes. Model inference was not run.

Signed commits

• Yes, I signed my commits. Human DCO attestation is required; the commit includes Assisted-by under repository policy.
• Documentation updated (docs/content/) for user-facing changes, or not applicable

[mudler-agent]

Closed: changes merged in batch PR #12221.