feat!: generate scores using static thresholding

[ab93]

Signed-off-by: Avik Basu ab93@users.noreply.github.com

• use payload factory with headers to instantiate the correct payload
• use numalogic v0.3.3
• generate scores using static threshold if no model is found
• changed the payload forwarding strategy - we now forward StreamPayload until threshold vtx with respective headers (MODEL_INFERENCE/STATIC_INFERENCE/MODEL_STALE) and in threshold vtx we conditional forward to postprocess and/or trainer vtx based on header

[codecov-commenter]

Codecov Report

Merging #80 (37736e8) into main (e289a90) will increase coverage by 0.47%. The diff coverage is 95.18%.

@@            Coverage Diff             @@
##             main      #80      +/-   ##
==========================================
+ Coverage   89.32%   89.79%   +0.47%     
==========================================
  Files          17       17              
  Lines         824      843      +19     
  Branches       74       76       +2     
==========================================
+ Hits          736      757      +21     
+ Misses         67       65       -2     
  Partials       21       21              

Impacted Files	Coverage Δ
numaprom/_constants.py	100.00% <ø> (ø)
numaprom/udsink/train_rollout.py	85.43% <25.00%> (+0.14%)	:arrow_up:
numaprom/__init__.py	83.33% <81.25%> (+6.41%)	:arrow_up:
numaprom/entities.py	95.34% <95.00%> (-1.58%)	:arrow_down:
numaprom/prometheus.py	66.19% <100.00%> (ø)
numaprom/redis.py	46.66% <100.00%> (ø)
numaprom/tools.py	88.37% <100.00%> (+1.30%)	:arrow_up:
numaprom/udf/filter.py	89.47% <100.00%> (ø)
numaprom/udf/inference.py	100.00% <100.00%> (ø)
numaprom/udf/postprocess.py	93.02% <100.00%> (ø)
... and 4 more

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