feat: ensemble scoring

[ab93]

• Enable ensemble scoring
• Add static_threshold_wt field in config
• Increase min training datapoints to 1000 -> 2000
• Increase threshold factor 3 -> 5 to make the model less sensitive
• Use TanhScaler as default preprocessing step

[codecov-commenter]

Codecov Report

Merging #92 (a7b4fb6) into main (9ab4a2b) will increase coverage by 0.27%. The diff coverage is 97.14%.

@@            Coverage Diff             @@
##             main      #92      +/-   ##
==========================================
+ Coverage   89.75%   90.03%   +0.27%     
==========================================
  Files          18       19       +1     
  Lines         898      933      +35     
  Branches       92       95       +3     
==========================================
+ Hits          806      840      +34     
- Misses         73       74       +1     
  Partials       19       19              

Impacted Files	Coverage Δ
numaprom/udf/window.py	94.91% <ø> (ø)
numaprom/udf/postprocess.py	96.42% <94.11%> (+0.04%)	:arrow_up:
numaprom/tools.py	86.33% <96.77%> (+1.06%)	:arrow_up:
numaprom/config/__init__.py	100.00% <100.00%> (ø)
numaprom/config/_config.py	100.00% <100.00%> (ø)
numaprom/entities.py	95.50% <100.00%> (+0.15%)	:arrow_up:
numaprom/udf/threshold.py	100.00% <100.00%> (ø)
numaprom/udsink/train_rollout.py	85.43% <100.00%> (ø)

:mega: We’re building smart automated test selection to slash your CI/CD build times. Learn more