Refactor `BasePlot` not to create a dataframe representation of the data

[ilan-gold]

What kind of feature would you like to request?

Additional function parameters / changed functionality / changed defaults?

Please describe your wishes

See #3717 for what prompted me to look at this code.

Currently BasePlot creates a in-memory copy as a dataframe of the main data of interest (obsm, X, layers etc.):

scanpy/src/scanpy/plotting/_baseplot_class.py

Lines 148 to 157 in 0b82c93

	self.categories, self.obs_tidy = _prepare_dataframe(
	adata,
	self.var_names,
	groupby,
	use_raw=use_raw,
	log=log,
	num_categories=num_categories,
	layer=layer,
	gene_symbols=gene_symbols,
	)

I believe this to be unnecessary as this dataframe is only ever used for groupby operations, for which we have a zero-copy solution in https://scanpy.readthedocs.io/en/latest/generated/scanpy.get.aggregate.html

Thus we should

• Refactor Baseplot not to create a copy
• Use the https://scanpy.readthedocs.io/en/latest/generated/scanpy.get.aggregate.html for aggregation
• Ensure this doesn't affect performance
• Integrate (feat): sc.get.aggregate via dask #3700

[dorzhey]

I'd like to work on this issue, will submit a PR soon.

[ilan-gold]

wooo great thanks @dorzhey !