Hardcoded spacing and missing timestamp if None in convert_to_xarray

[robert-DL]

Hi folks,

in the era5 dataset, the spacing of lats and longs is hardcoded, leading to problems if people use other resolutions than 240x121 and want to store predictions.

https://github.com/INRIA/geoarches/blob/66cd870fdcf737d7bece759458d4aee05f0c0d0a/geoarches/dataloaders/era5.py#L179C9-L196C10

Additionally, the timestamp parameter in convert_trajectory_to_xarray is set to None, but if you do not pass it, convert_to_xarray receives None and can_t handle that as None objects have no cpu() property.
See here:

geoarches/geoarches/dataloaders/era5.py

Line 178 in 66cd870

times = pd.to_datetime(timestamp.cpu().numpy(), unit="s").tz_localize(None)

Best
Robert