[JS API] Add import_model(tensor) to Node.js API

[Nishant-ZFYII]

Summary

Expose Core::import_model(tensor, device, config) to the Node.js API so compiled models can be imported from an ov.Tensor, matching the Python bindings (#32651).

Fixes #32725

What/Why

OpenVINO C++ already supports importing a compiled model from a Tensor (#31327). This PR wires that overload into the JavaScript/Node.js bindings so users can round-trip a compiled model via exportModelSync() → Tensor → importModelSync() (and the async equivalents).

Changes

• Add a Tensor overload for CoreWrap::import_model() (sync)
• Add a Tensor overload for CoreWrap::import_model_async() (async)
• Update TypeScript definitions to accept Tensor as the first parameter
• Add unit tests covering Tensor import for both sync and async paths

Testing

• All existing and new tests pass locally
• Implementation follows existing patterns and reuses helper utilities (validate_value, cast_to_tensor, to_anyMap)

Example

const { addon: ov } = require('openvino-node');
const core = new ov.Core();

// Export compiled model to a buffer
const compiled = core.compileModelSync(model, 'CPU');
const buffer = compiled.exportModelSync();

// Wrap bytes in a Tensor
const uint8Array = new Uint8Array(buffer);
const tensor = new ov.Tensor(ov.element.u8, [buffer.length], uint8Array);

// Import compiled model from Tensor (NEW)
const imported = core.importModelSync(tensor, 'CPU');

## Related PRs
- C++ support: #31327
- Python parity: #32651