make calc_batch_size compatible with default__load_csv_rows

[dataders]

dbt-labs/dbt-core#3623 (released as part of dbt-core v0.21.0) allowed adapters like dbt-sqlserver to not have to re-implement default__load_csv_rows() because the only difference between dbt-sqlserver's and core's global_project was the binding parameter character: ?, and %s respectively.

What I had forgotten about was @jacobm001's calc_batch_size which doesn't really work with default__load_csv_rows.

It'd be nice to not have to reimplement default__load_csv_rows! This might require a PR to dbt-core directly...

[jacobm001]

Should be a pretty easy thing to migrate into the typical dbt macros. It'll probably be a couple days before I can get to it, but I can get that done.

[sdebruyn]

I worked a bit on #211 and made sure it followed the core implementation more closely. The macro to get the batch size in dbt-core doesn't have the model or its columns as argument and we need that to calculate our batch size based on the amount of columns the csv contains. Changing the macro in dbt-core would probably cause incompatibility with adapters which haven't updated yet.

I'd be glad to contribute to dbt-core, but I'd need some advice on how to change the get_batch_size() in dbt-core so that it takes an argument. And which argument would that be? The model, the columns or the amount of columns?

@dataders @jtcohen6 ideas?