Add optional ts into AbstractPipeline.forecast

[Mr-Geekman]

🚀 Feature Request

Add optional parameter ts into AbstractPipeline.forecast with default value None. It will allow users to use new TSDatasets after loading the pipeline.

Proposal

1. Add optional parameter ts into AbstractPipeline.forecast with default value None. It will allow users to use new TSDatasets after loading the pipeline.
2. Update children.
3. Change the signature in all the children.
4. Add logic for checking if ts is given:
   * If ts is given then use ts, ignore its transforms during making the forecasting.
   * If ts isn't given then use self.ts, if it isn' present raise the error.
5. In notebook with inference demo add example with making forecast on a new ts without passing it during load.

Test cases

1. Make sure current tests pass.
2. For every pipeline check.

• That it is possible to make forecast on other ts. For example, keep only fraction of segments for check.
• That it is possible to load pipeline from disk without ts and make forecast by passing a new ts.

Additional context

No response

[Mr-Geekman]

Closed by #1071.