feat(gallery): add MiniCPM-V 4.7 1B variants

[localai-org-maint-bot]

Description

Add MiniCPM-V 4.7 1B for text chat and image understanding. The gallery currently stops at MiniCPM-V 4.6. Offer the official Q4_K_M, Q5_K_M, Q6_K, and Q8_0 GGUF builds as variants, each with the F16 vision projector and an 8,192-token default context. Document installation and explicit variant selection.

Notes for Reviewers

• Sources: original model and official GGUF builds.
• Downloads are pinned to HF revision 59cd286146b6ce70f818508ccf512e11438d8957. All five SHA256 values independently match HF API LFS metadata and the pinned download URLs' x-linked-etag headers.
• GGUF headers identify qwen35 weights and the minicpmv4_7 projector. Both are supported by LocalAI's pinned llama.cpp revision: architecture registry and projector loader. No backend change is needed.
• Reuse the existing Qwen template, including the embedded tokenizer template and Jinja runtime.
• Standalone Go/yaml.v3 validation passes: 2,195 unique model names, valid variant targets, four additions, and all 2,191 existing entries unchanged. Model/projector paths, pinned hashes, backend, context, use cases, and template settings pass validation.
• git diff --check passes.
• go test ./core/gallery cannot start because this checkout lacks generated pkg/grpc/proto bindings. No inference test was run.

Signed commits

• Human DCO sign-off required. The commit contains Assisted-by: Codex:gpt-6; repository policy prohibits AI sign-off.
• Documentation updated under docs/content/.