fix: Fix read_csv to respect the order specified by the columns argument

[romanovacca]

fix #13066

in fn parse_csv(), we use get_projection(), which uses the projection that contains the correct columns and order. That function mentions:
we also need to sort the projection to have predictable output, which happens via sort_unstable(), which leads to a different order of the projection/columns.

In my fix, we now take the original projection and use that to fetch the matching column names, and when all the chunks of data are processed, I reorder the dataframe (if column_names was specified), into the order that was initially passed.

I would expect this reordering to be present somewhere already, but I couldn't find it

[stinodego]

I did a rebase. However, I don't think this is the right fix. The offending code seems to be here:

polars/crates/polars-io/src/csv/read_impl/mod.rs

Lines 227 to 240 in 11c6f9b

	if let Some(cols) = columns {
	let mut prj = Vec::with_capacity(cols.len());
	for col in cols {
	let i = schema.try_index_of(&col)?;
	prj.push(i);
	}
	// update null values with projection
	if let Some(nv) = null_values.as_mut() {
	nv.apply_projection(&prj);
	}
	projection = Some(prj);
	}

After this, the columns information is lost. So we should make sure the projection is correct at that point, rather than try to fix it later.

Since this PR is pretty old and has been on draft for a while, and it doesn't seem (close to) ready, I'll close it for now. Feel free to open a fresh PR and continue the work!