Correction of yield_loss_perc calculation and associated I/O

[chpolste]

Following up on point 4 raised in #22, a proposed fix for writing the yield_loss_perc values.

• Verify that csv and netcdf outputs are identical.

[chpolste]

Output consistency tested by running

import numpy as np
import pandas as pd
import xarray as xr

yield_loss_npy = np.load('agriculture_workflow/results/Cataluña_yield_loss_NUMPY.npy')
yield_loss_csv = pd.read_csv('agriculture_workflow/results/Cataluña_yield_loss_SPREADSHEET.csv')
yield_loss_nc = xr.open_dataset('agriculture_workflow/results/agridrought_hazard_ES51_mpi_m_mpi_esm_lr-smhi_rca4_rcp26_2046-2050.nc')

np.testing.assert_allclose(
    yield_loss_npy,
    yield_loss_csv[["maize", "wheat"]].values.reshape(yield_loss_npy.shape)
)

np.testing.assert_allclose(
    yield_loss_npy,
    yield_loss_nc["yield_loss"].transpose("y", "x", "crop").values
)

with the default configuration of the workflow.