Gluonts

Backporting fixes:
- Fix iterable.Cached. 3060 by jaheba
- Torch: Remove double caching of dataset. 3061 by jaheba

Backporting fixes:
- Fix edge cases in metric computation 3037 (lostella)
- Refactor tests for ev.aggregations 3038 (lostella)
- Add plt.show() to README 3043 (lostella)
- Rotbaum: Add item-id to forecast. 3049 (jaheba)
- Fix mypy checks 3052 (lostella)

See full diff here: https://github.com/awslabs/gluonts/compare/v0.13.0...v0.14.0

Breaking changes

* Simplify plotting of forecasts. [2864](https://github.com/awslabs/gluonts/pull/2864)
* Breaking: change num_output with quantiles in TFT (mxnet) [2879](https://github.com/awslabs/gluonts/pull/2879)
* DeepVARHierarchicalEstimator: Remove target_dim parameter. [2889](https://github.com/awslabs/gluonts/pull/2889)
* Rework maybe.Maybe to use a boxed container. [2900](https://github.com/awslabs/gluonts/pull/2900)
* Rework ev. [2914](https://github.com/awslabs/gluonts/pull/2914)
* gluonts.ev: Rework exports. [2928](https://github.com/awslabs/gluonts/pull/2928)
* Remove freq argument in SeasonalNaivePredictor [2932](https://github.com/awslabs/gluonts/pull/2932)
* Remove freq argument from naive2 [2935](https://github.com/awslabs/gluonts/pull/2935)
* Simplify PyTorchPredictor serde [2965](https://github.com/awslabs/gluonts/pull/2965)
* Move QuantizeScaled from Wavenet to gluonts.transform, rename to QuantizeMeanScaled [2975](https://github.com/awslabs/gluonts/pull/2975)
* Change torch state_dict file name to prediction-net-state.pt [2985](https://github.com/awslabs/gluonts/pull/2985)
* Remove freq attribute from NPTS [3009](https://github.com/awslabs/gluonts/pull/3009)

Backporting:
- Fix Rotbaum serialization and deserialization 3068 by pantanurag555
- Fix Rotbaum to handle short series 3073 by leica2023