Improve heatmap loading time by using pandas (or even better : polars) library ?

https://pola.rs/ or https://pandas.pydata.org/

• fix y dimension ordering
• return np.array for y_values ? no difference
• fix result computation slowness (vectorize)
• investigate why figure method is slower when using pandas
• vectorize everything (no more apply() method)?
  • after groupby in build_z_values()
  • in build_y_values()
• specify dtypes
• use iterator() ? slower with iterator()