🐛 fix overriding table's short name

[Marigold]

The problem is in these lines

# Read economy data.
tb_economy = pr.read_csv(z.open(DATA_PATH_ECONOMY), metadata=metadata, origin=origin)
tb_economy.metadata.short_name = "emissions_weighted_carbon_price_economy"

# Read coverage data.
tb_coverage = pr.read_csv(z.open(DATA_PATH_COVERAGE), metadata=metadata, origin=origin)
tb_coverage.metadata.short_name = "emissions_weighted_carbon_price_coverage"

metadata refers to the same object and the second tb_coverage.metadata.short_name updates the short name of the first table, too. Function create_dataset, which accepts two tables, then overwrites the tables.

I've added a uniqueness check for table short names and also fixed this by copying metadata object inside pd.read_csv function.

(Found this bug in our full rebuild.)

[pabloarosado]

Looks good! The only thing I don't understand is why this happened now, and not when the step was created, less than two months ago.

[Marigold]

That's a good question. I don't know either. It could be related to the plog changes (I were doing a lot of changes to tables.py) or it could be that the table somehow stayed cached.