Fix the PTQ NaN bug

Description

After the PTQ, some architectures were producing NaN outputs when benched.

It was caused by a datatype issue during the CLE step : values were manipulated as floats instead of doubles.

Note that the upcoming improvements branch will completely rework the handling of the data types.

Modified files : CLE.cpp and CLE.hpp

include/aidge/quantization/PTQ/CLE.hpp

@@ -30,7 +30,7 @@ namespace Aidge
      * @param graphView The GraphView to process.
      * @param targetDelta the stopping criterion (typical value : 0.01)
      */
-    void crossLayerEqualization(std::shared_ptr<GraphView> graphView, float targetDelta = 0.01);
+    void crossLayerEqualization(std::shared_ptr<GraphView> graphView, double targetDelta = 0.01);
 
 }