feat: add gift eval module

[AzulGarza]

this pr adds the gift_eval module. it allows to evaluate time copilot forecasters in a reproducible and easy way using the gift eval tools. it includes ci tests showing that the results obtained by the seasonal naive model are the exact ones as the provided by the mantainers in hugging face

[Copilot]

Pull Request Overview

This PR introduces a new gift_eval module to TimesCopilot that enables reproducible evaluation of time series forecasters using the GIFT-Eval benchmark framework. The module integrates with GluonTS to provide standardized evaluation metrics and comparison against established baselines like seasonal naive models.

• Adds core evaluation infrastructure with GIFTEval class and GluonTSPredictor adapter
• Implements dataset loading and preprocessing utilities for GIFT-Eval compatible datasets
• Includes comprehensive CI tests that validate results match official seasonal naive benchmarks

Reviewed Changes

Copilot reviewed 10 out of 11 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
timecopilot/gift_eval/utils.py	Defines dataset lists and configurations for short/medium/long term evaluations
timecopilot/gift_eval/gluonts_predictor.py	Adapter class to wrap TimeCopilot forecasters as GluonTS predictors
timecopilot/gift_eval/eval.py	Main evaluation class with predictor evaluation and CSV output functionality
timecopilot/gift_eval/data.py	Dataset loading and preprocessing utilities adapted from GIFT-Eval
tests/gift_eval/test_evaluation.py	Test suite validating evaluation results against official benchmarks
tests/gift_eval/conftest.py	Test fixtures for dataset downloading and caching
pyproject.toml	Adds gift_eval test marker configuration
mkdocs.yml	Adds gift_eval module to documentation navigation
docs/api/gift_eval/gift_eval.md	Documentation page for the gift_eval module
.github/workflows/ci.yaml	Adds CI workflow for gift_eval tests

[Copilot]

[nitpick] The print statement should use a logger instead of print() for better control over output in production environments.

Suggested change

	print(
	logging.info(

[AzulGarza]

thanks, fixed in AzulGarza/timecopilot@944fd8a.

[Copilot]

The hardcoded URL should be extracted to a module-level constant or configuration variable for better maintainability.

Suggested change

	"https://raw.githubusercontent.com/SalesforceAIResearch/gift-eval/refs/heads/main/notebooks/dataset_properties.json"
	DATASET_PROPERTIES_URL

[AzulGarza]

thanks, fixed in AzulGarza/timecopilot@944fd8a.