Fix SOI uprating dtype error (#602)

* Fix SOI dtype error and ensure non-negative retirement/SS values

1. soi.py: Cast SOI "Value" column to float before uprating multiplication,
   fixing TypeError on newer pandas that reject float-into-int64 assignment.

2. extended_cps.py: Add defensive np.maximum(pred_values, 0) clip in
   _splice_cps_only_predictions for retirement contribution and SS
   sub-component variables, preventing negative values from surviving
   entity-level aggregation.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Remove defensive clip — apply_retirement_constraints already handles this

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Author: MaxGhenis

changelog.d/fix-soi-dtype-and-se-pension.fixed.md

@@ -0,0 +1 @@
+Fix SOI uprating dtype error on newer pandas and add defensive non-negativity clip for retirement/SS variables in splice step.

policyengine_us_data/utils/soi.py

@@ -164,6 +164,7 @@ def get_soi(year: int) -> pd.DataFrame:
     }
     soi = pd.read_csv(CALIBRATION_FOLDER / "soi_targets.csv")
     soi = soi[soi.Year == soi.Year.max()]
+    soi["Value"] = soi["Value"].astype(float)
 
     uprating_factors = {}
     for variable in uprating_map: