Rx2day, Rx5day resulting in some infinite values

Two extreme precipitation indices, rx2day and rx5day, produces some infinite values on e.g. GridClim and PTHBV datasets. For both these datasets there are no issues with rx1day.

Example (on bi):

import iris

from climix.metadata import load_metadata
from dask.distributed import Client

client = Client()

fname = "/nobackup/rossby26/users/sm_erhol/extremeEventAttribution/gavle2021/pr_gavle2021_SMHIGridClim_day_19610101-20181230.nc"
# This will be an masked array
cube = iris.load_cube(fname)

index_catalogue = load_metadata()
index = index_catalogue.prepare_indices(["rx2day"])[0]

index_cube = index([cube], client)

assert not np.any(np.isinf(index_cube.data))

When realising the index_cube some overflow warnings are thrown. If the spatial dimensions of the cube are collapsed (average, max), no infinite values are produced.