Model different queries to the data source with Query

[Jaapel]

A large part of what resolvers, drivers and data_adapters do is querying the data (assumed to be large) data source, so that only a small part of that data source needs to be present. That happens at discovery-time (e.g. do not read all files, but only read the relevant files), during parsing of the file (e.g. doing an http-range request to only read certain bytes of the COG), or during post-processing (data adapter).

What can be queried depends on the data type. A geodataset has no zoom level. A dataset has no geometry component. Currenly, there is a lot of boilerplate code, as different DataSources, have different keyword only parameters that differ slightly from other datasets.
If all classes would accept a different object, instead of all these different parameters, a lot of HydroMT logic could be better generalized.
These query parameters could still exist in higher-level methods (DataCatalog.get_rasterdataset), but would contribute to deduplication efforts in the DataSource, DataAdapter and Driver classes.