fix: implicit conversion to float64

There was an implicit conversion to float64 taking place due to a cast
to the Python's list (and hence Python's float). This resultued in a
serious (fator ~10) degradation in performance, which should now be
fixed.

The performance degradation was a result of temporary allocation
performed by pandas when a dtype of a frame was implicitly changed
in updates of the form e.g.:
```python
import pandas as pd
import numpy as np

df = pd.DataFrame({"f32": np.ones(2, dtype="float32")})
df.iloc[1:2] = np.float64(1/3)
```
since pandas 2.1, such operations raise a FutureWarning. All occurences
of that warning in DEMENTpy are resolved in this commit.

Author: Mikolaj-A-Kowalski

src/grid.py

@@ -420,13 +420,13 @@ def metabolism(self,day):
         # Update Substrates pools with dead enzymes
         DeadEnz_df = pd.concat(
             [Enzyme_Loss,
-            Enzyme_Loss.mul(self.Enz_Attrib['N_cost'].tolist()*self.gridsize,axis=0),
-            Enzyme_Loss.mul(self.Enz_Attrib['P_cost'].tolist()*self.gridsize,axis=0)],
+            Enzyme_Loss.mul(np.repeat(self.Enz_Attrib['N_cost'].values, self.gridsize), axis=0),  # Implicit f64 conversion
+            Enzyme_Loss.mul(np.repeat(self.Enz_Attrib['P_cost'].values, self.gridsize), axis=0)], # Implicit f64 conversion
             axis=1
         )
         DeadEnz_df.index = [np.arange(self.gridsize).repeat(self.n_enzymes), DeadEnz_df.index] # create a multi-index
         DeadEnz_gridcell = DeadEnz_df.groupby(level=0).sum()  # total dead mass across taxa in each grid cell
-        self.Substrates.loc[is_deadEnz] += DeadEnz_gridcell.values
+        self.Substrates.loc[is_deadEnz] += DeadEnz_gridcell.values 
 
 
     def mortality(self,day):       
@@ -713,5 +713,5 @@ def reinitialization(self,initialization,microbes_pp,output,mode,pulse,*args):
         # last: assign microbes to each grid box randomly based on prior densities
         choose_taxa = np.zeros((self.n_taxa,self.gridsize), dtype='int8')
         for i in range(self.n_taxa):
-            choose_taxa[i,:] = np.random.choice([1,0], self.gridsize, replace=True, p=[frequencies[i], 1-frequencies[i]])
+            choose_taxa[i,:] = np.random.binomial(1, frequencies[i], self.gridsize)
         self.Microbes.loc[np.ravel(choose_taxa,order='F')==0] = np.float32(0) # NOTE order='F'