[Bug] - River.from_imod5_data unacceptably slow for IBRAHYM iMOD5 test

[JoerivanEngelen]

Bug description
The IBRAHYM imod5 test hangs on calling River.from_imod5_data. The example has 37 timesteps for RIV and 37 layers, and need to go through quite some processing steps. There are 2 systems, which both need to be allocated to layers. Interesting is that the Drain package has 4 systems, 3 of which need to be allocated to layers and 37 layers as well. This one is slow, but not as unbearable as the River.

I have two hypotheses:

1. The task graph becomes too complex for this package. This might be fixed in a similar manner as [Bug] Big performance regression on idf.open_subdomains #1844. What speaks against this is that the input grids only contain one layer.
2. Too many layers, caused by [FEATURE] - Allocate RIV, DRN, RCH: drop unused layers #1821. What speaks against this is that the data should not be even loaded into memory yet.