Feature: add toy data

[jsmariegaard]

>>> import modelskill as ms
>>> cc = ms.data.vistula()
>>> cc

<ComparerCollection>
Comparers:
0: Tczew - Discharge [m3/s]
1: Krasnystaw - Discharge [m3/s]
2: Sandomierz - Discharge [m3/s]
3: Szczucin - Discharge [m3/s]
4: Nowy Sacz - Discharge [m3/s]
5: Tryncza - Discharge [m3/s]
6: Ptaki - Discharge [m3/s]
7: Suraz - Discharge [m3/s]

>>> cc = ms.data.oresund()
>>> cc

<ComparerCollection>
Comparers:
0: Drogden - Surface Elevation [meter]
1: Barseback - Surface Elevation [meter]
2: Helsingborg - Surface Elevation [meter]
3: Kobenhavn - Surface Elevation [meter]
4: Koege - Surface Elevation [meter]
5: MalmoHamn - Surface Elevation [meter]
6: Vedbaek - Surface Elevation [meter]

Both datasets are now around 1MB, both contains aux data and attrs, that could be used for examples and testing in the future (not yet used). The data module has been added to the api docs.

And a later point it would be great to add more datasets: nortseawaves, ...

It would also be great to add the new notebook to the examples in docs.

[jsmariegaard]

What is an acceptable data file size to include in the package? 1MB per case? Should we maybe remove some stations in above examples? Could we reduce to float16 or use other tricks to save some file size?

[ecomodeller]

I think we can remove some stations and change to float32.

[jsmariegaard]

Ways to make datasets smaller on disk:

• reduce time period
• reduce number of observations
• float32 instead of float64
• crop modelresult to period covered by obs
• reduce time resolution of modelresults (e.g. 3 hourly instead of 30 min)

Further ideas to reduce disk size (not tried):

• use compression via https://docs.xarray.dev/en/stable/generated/xarray.Dataset.to_netcdf.html
• store data in alternative format e.g. store all measurements in a single parquet file and keep additional metadata in the function definition for each datasets