feat(gallery): add Hy-MT2 7B GGUF variants

[localai-org-maint-bot]

Description

Add Hy-MT2-7B Q4_K_M, Q6_K, and Q8_0 builds to the model gallery, linked as variants. This fills the gap beside the existing 1.8B translation model. Include installation commands and a translation prompt example.

Notes for Reviewers

• Official weights and original model; Apache-2.0.
• All three downloads are pinned to ab8472660ac61fac25f1af43fac2599d52a8a775. Each SHA256 matches both HF LFS metadata and the pinned download's x-linked-etag header. Preserve the publisher's different Hy/HY filename casing.
• HF metadata identifies hunyuan-dense, supported by LocalAI's pinned llama.cpp revision 50631b3d2c569ad8e5c112090cd28570b1268ee0. These are standard K-quants/Q8_0, not the separate STQ formats.
• Use native Jinja chat formatting, publisher sampling defaults, and an 8K initial context. Users can increase context within the model's 262K limit.
• go test ./core/gallery -count=1 -ginkgo.focus='gallery/index.yaml|gallery variant lint helpers' passes. A semantic YAML comparison confirms all 1,890 existing entries remain unchanged. All new model paths and hashes match pinned HF metadata. git diff --check passes.
• Full go test ./core/gallery -count=1: 392 passed, 3 failed, 9 skipped, identically on unchanged master and this branch. Failures are Forbidden responses from external GitHub/gist fixtures at request_test.go:22, backends_test.go:208, and backends_test.go:213.
• Inference was not run locally.

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