Pipelinets

- New feature: Ability to define custom names for models in PipelineConfigs class
- New feature: When ModelPipeline is instantiated, the compute device name is automatically printed.
- Improved functionality: Modify the default learning_rate of time2vec model.
- Fixed bug: Issue with attribute error in PipelineConfigs parameter, where the function name incorrectly displayed as '\_\_init\_\_'.
- Fixed bug: Disable gbdt models training log printing when training with ModelPipeline.

- Added support for using the "model_name__`init_params`" as the init parameter, allowing separate specification of initialization parameters for submodels in the ModelPipeline.
- Changed the device parameter of the spinesTS model to accelerator, and changed its default value from cpu to auto.
- Increased the batch_size of the spinesTS model to improve training speed.

Fixed a bug in 0.3.6 where darts models could not be saved when saving ModelPipeline class object.

Lots of new features added, documentation will be improved in the future, stay tuned