[ENH] Add predict to v2 models

[phoeenniixx]

This PRs adds the predict function and related functionality to v2
Fixes #1883

• Add PredictCallback
• Add Base_pkg class and predict wrapper
• Add tests for predict
• Add tests for checkpointing
• Add support of yml file reading for cfgs (similar to pytorch-tabular)

[codecov]

Codecov Report

❌ Patch coverage is 81.13208% with 50 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (main@d4596de). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
pytorch_forecasting/base/_base_pkg.py	76.23%	24 Missing ⚠️
pytorch_forecasting/callbacks/predict.py	64.28%	20 Missing ⚠️
pytorch_forecasting/models/base/_base_model_v2.py	77.27%	5 Missing ⚠️
...ytorch_forecasting/tests/test_all_estimators_v2.py	96.55%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1984   +/-   ##
=======================================
  Coverage        ?   86.17%           
=======================================
  Files           ?      162           
  Lines           ?     9600           
  Branches        ?        0           
=======================================
  Hits            ?     8273           
  Misses          ?     1327           
  Partials        ?        0           

Flag	Coverage Δ
cpu	86.17% <81.13%> (?)
pytest	86.17% <81.13%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[fkiraly]

minor formatting issue: please have newlines around bullet point lists

[fkiraly]

Extremely neat!

I like the design, and I think this is brilliantly done!

Some questions and requests:

• could you write a summary for the PR?
• could you add a vignette somewhere in a notebook or an "examples" section? I think this is crucial to have a full review.
• can you explain how or why the test configs can be replaced by a single test vignette?
• design questions: why are fit and predict kwargs there and not in __init__? Have you thought about this?