[Major] Compute tabular data samples at batch creation [WIP]

[ourownstory]

🔬 Background

• Currently very high data consumption when loading global models with many IDs. Would fix Drastically increased memory load #1469

🔮 Key changes

• On the fly sampling of the dataloader

📋 Review Checklist

• I have performed a self-review of my own code.
• I have commented my code, added docstrings and data types to function definitions.
• I have added pytests to check whether my feature / fix works.

[codecov]

Codecov Report

Attention: 25 lines in your changes are missing coverage. Please review.

Comparison is base (305de42) 88.41% compared to head (a6eceb2) 88.13%.

Files	Patch %	Lines
neuralprophet/time_dataset.py	93.24%	20 Missing ⚠️
neuralprophet/forecaster.py	50.00%	5 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1510      +/-   ##
==========================================
- Coverage   88.41%   88.13%   -0.29%     
==========================================
  Files          38       38              
  Lines        5103     5115      +12     
==========================================
- Hits         4512     4508       -4     
- Misses        591      607      +16     

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