QVAC-24975 feat[translation-nmtcpp]: align with main-gpu selection contract

[Alok-Ranjan23]

🎯 What problem does this PR solve?

• translation-nmtcpp had its own bespoke GPU-selection surface, out of step with the shared main-gpu contract introduced across the other ggml/whisper addons under QVAC-24975.

📝 How does it solve it?

• Accepts main-gpu (and case-alternative main_gpu alias — exactly one) as either a raw ggml registry index (number or integer string) or "dedicated" / "integrated" (case-insensitive).
• Rejects combining the new selector with the pre-existing gpu_backend/gpu_device (or camelCase aliases).
• Raw in-range index preserves its registry identity through backend/safety filtering; excluded targets fall back to CPU. Out-of-range indices warn and use the default dedicated-first selection. Class selectors are strict — unavailable classes fall back to CPU.
• Adds the surface to the JS wrapper, native forwarding, AddonJs, TypeScript types, README, and consumer type tests.

🧪 How was it tested?

• packages/translation-nmtcpp/addon/tests/nmt_gpu_selection_test.cpp + additions to translation_tests.cpp cover the new native selector paths.
• packages/translation-nmtcpp/test/unit/wrapper-lifecycle.test.js exercises the JS validation branches.
• packages/translation-nmtcpp/test/types/consumer-cjs.test-d.ts type-tests both the widened number | string accepted type and the @ts-expect-error rejection surface.

🔌 API Changes

```typescript
new TranslationNmtcpp({
modelType: "IndicTrans",
"main-gpu": "dedicated",
});

new TranslationNmtcpp({
modelType: "IndicTrans",
main_gpu: 1,
});
```

Refs: QVAC-24975.

[github-actions]

Review Status

Current Status: ❌ PENDING
Approvals so far: none

Pending reviews: Needs 1 Management or Team Lead, and 1 more from Management, Team Lead, or Member.

[github-actions]

License compliance — clean

No new dependency license findings in this PR.

Warn-only (shadow) mode — this check does not block merges yet.

Updated automatically by the canonical license compliance workflow.

NOTICE presence (advisory)

Missing NOTICE (advisory, does not block):

• ./docs/website
• ./packages/fabric/test/integration
• ./packages/llm-llamacpp/benchmarks/server
• ./packages/llm-llamacpp/benchmarks/performance
• ./packages/inference-addon-cpp/mobile
• ./packages/asr-ggml/benchmarks/server
• ./packages/embed-llamacpp/benchmarks/server
• ./packages/embed-llamacpp/benchmarks/performance
• ./packages/sdk/e2e
• ./packages/vla-ggml/sim/server
• ./.github/actions/release-merge-guard

[GustavoA1604]

[medium] packages/translation-nmtcpp/addon/src/model-interface/TranslationModel.cpp — line 719 (if (selector != config.end()) {)

mainGpu_ is only updated when the replacement config contains a selector. Calling setConfig({use_gpu: 1, main-gpu: "dedicated"}), loading, then
replacing it with setConfig({use_gpu: 1}) leaves "dedicated" active and can incorrectly fall back to CPU on an integrated-only host. Reset
mainGpu_ and its selection state when the incoming map omits the selector, and add a reconfiguration regression test.

[low][qvac-packages] packages/translation-nmtcpp/src/index.ts — line 753 ("main-gpu"?: number | string;)

This changes the published NPM API and device-selection behavior, but packages/translation-nmtcpp/CHANGELOG.md has no [Unreleased] entry. Add a
concise consumer-facing entry describing the selector and its fallback/conflict behavior.