feat(gallery): add Granite 4.2 variants

[localai-org-maint-bot]

Description

Adds the official IBM Granite 4.2 GGUF models in 3B, 8B, and 30B sizes. Each size includes Q4_K_M and Q8_0 builds connected through gallery variants.

The entries use llama.cpp with the embedded Jinja tokenizer template for multilingual chat, reasoning, coding, and tool use.

Verification

• Confirmed all six SHA256 values against Hugging Face x-linked-etag
• Confirmed all model and artifact URLs return successfully
• Confirmed Granite 4.2 uses the dense Granite architecture supported by the pinned llama.cpp revision
• git diff --check https-origin/master...HEAD
• Independent gallery review passed
• go test ./core/gallery could not start locally because generated pkg/grpc/proto sources are absent; make protogen-go was blocked when the environment proxy rejected the protoc download

Notes for Reviewers

CI should provide the generated protobuf prerequisite and run the gallery suite.

Signed commits

• Yes, I signed my commits.
• Documentation updated (docs/content/) for user-facing changes, or not applicable