Blocked by #6715
Blocked by #7281

After #7281 goes in, most (maybe all) IO operations in the dataframe and array modules can be converted into Blockwise high-level graph (HLG) layers. This change should improve HLG optimizations (like blockwise fusion).

This issue is intended as a "living" checklist of action items related to Blockwise IO...

Dask-Dataframe:

• read_parquet [Use Blockwise for DataFrame IO (parquet, csv, and orc) #7415]
• read_csv (via text_blocks_to_pandas) [Use Blockwise for DataFrame IO (parquet, csv, and orc) #7415]
• read_orc [Use Blockwise for DataFrame IO (parquet, csv, and orc) #7415]
• make_timeseries [Move timeseries and daily-stock to Blockwise #7615]
• daily_stock [Move timeseries and daily-stock to Blockwise #7615]
• read_hdf [Move read_hdf to Blockwise #7625]
• from_array
• from_pandas
• from_bcolz
• from_dask_array
• from_delayed
• read_json

Dask-Array:

• from_array
• from_zarr (may be covered by from_array)
• from_delayed
• from_func
• empty_like
• ones_like [Prepare Blockwise + IO infrastructure #7281]
• zeros_like [Prepare Blockwise + IO infrastructure #7281]
• full_like [Prepare Blockwise + IO infrastructure #7281]
• linspace
• arange
• eye
• diag
• diagonal
• triu
• tril
• fromfunction [Rewrite da.fromfunction w/ da.blockwise #7704]
• repeat