[WIP] Add tslearn.foundation module for re-using pre-trained time series models (typically from HF)

[rtavenar]

Introduces ZeroShotForecaster, LinearProbeForecaster, LinearProbeClassifier and the underlying TimeSeriesFoundationEmbedder, letting pre-trained models (e.g. Chronos-2) be used behind the usual tslearn estimator API, either zero-shot or with a linear head fitted on frozen representations.

[codecov]

Codecov Report

❌ Patch coverage is 99.51885% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 95.75%. Comparing base (4cf2840) to head (96faf83).

Files with missing lines	Patch %	Lines
tests/test_foundation.py	99.56%	3 Missing ⚠️
tslearn/foundation/_embedding.py	98.70%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #699      +/-   ##
==========================================
+ Coverage   95.14%   95.75%   +0.60%     
==========================================
  Files          83       88       +5     
  Lines        7814     9061    +1247     
==========================================
+ Hits         7435     8676    +1241     
- Misses        379      385       +6     

☔ View full report in Codecov by Harness.
📢 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.