Consider adding index on file names to BMI config, hydrofabric, and possibly other datasets

[robertbartel]

Initially, it seemed too tedious to, e.g., define and constrain a dataset or a data requirement for BMI data on the specific file names available/needed. However, with the addition of automatic domain detection, it is worth considering whether this would be worth having. It would allow for more direct aligning of files to needs (in a realization config).

There are other datasets that may also benefit. Hydrofabric datasets in particular could use file constraints to indicate whether attribute data is available for use with BMI init config generation. CSV forcing datasets might also could use this.

However, aside from making domains much more complex, it would also make certain things tricky. E.g., a realization config dataset implies certain requirements about what forcing file names and what BMI file names are. These are requirements of a job that uses that realization config dataset, though. They are not aspects of the domain of the realization config dataset (not in the sense of constraining the files present within the dataset).

It should be carefully considered whether it would be beneficial, in practice, to implement these ideas in various places, and, if it is, we should do so.