Add support for PatchTST (time-series-forecasting)

[xenova]

This PR adds support for PatchTST models (time-series-forecasting).

ONNX export support is being added here: huggingface/optimum#2101

Export example:

git clone https://github.com/huggingface/transformers.js.git
cd transformers.js
pip install -q -r scripts/requirements.txt
pip install --upgrade git+https://github.com/huggingface/optimum@add-patchtst-onnx

followed by

python -m scripts.convert --quantize --model_id ibm/test-patchtst

---

Closes #1047

Example usage:

import { PatchTSTForPrediction, Tensor } from '@huggingface/transformers';

const model_id = "onnx-community/test-patchtst";
const model = await PatchTSTForPrediction.from_pretrained(model_id, { dtype: "fp32" });

// Example input
const dims = [64, 512, 7];
const prod = dims.reduce((a, b) => a * b, 1);
const past_values = new Tensor('float32',
    Float32Array.from({ length: prod }, (_, i) => i / prod),
    dims,
);
const { prediction_outputs } = await model({ past_values });
console.log(prediction_outputs);

outputs match pytorch within 1e-5 atol.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[xenova]

Example usage for PatchTSMixerForPrediction:

import { PatchTSMixerForPrediction, Tensor } from '@huggingface/transformers';

const model_id = "onnx-community/granite-timeseries-patchtsmixer";
const model = await PatchTSMixerForPrediction.from_pretrained(model_id, { dtype: "fp32" });

const dims = [64, 512, 7];
const prod = dims.reduce((a, b) => a * b, 1);
const past_values = new Tensor('float32',
    Float32Array.from({ length: prod }, (_, i) => i / prod),
    dims,
);
const { prediction_outputs } = await model({ past_values });
console.log(prediction_outputs);