Experiment with vendoring cpp11

[chainsawriot]

At least for now, it sounds like a decent idea to try and see the potential performance change.

ref

• Found non-API calls to R: SETLENGTH, SET_TRUELENGTH r-lib/cpp11#355
• fix: Avoid R CMD check warning regarding SETLENGTH() and SET_TRUELENGTH() duckdb/duckdb-r#145

[chainsawriot]

The experience of submitting minty to CRAN is still troubling me.

[chainsawriot]

We do not grow any cpp11 vector at all, so we should expect no performance impact.

readODS/src/read_ods_internals.cpp

Lines 231 to 252 in 60a1bcd

	cpp11::writable::strings cell_values(out_width*out_length + 2);
	cell_values[0] = std::to_string(out_width);
	cell_values[1] = std::to_string(out_length);
	int t = 2;
	for (unsigned int i = 0; i < contents.size(); i++){
	for (unsigned int j = 0; j < contents[i].size(); j++){
	cell_values[t] = (contents[i][j] != 0) ?
	Rf_mkCharCE(parse_single_cell(contents[i][j], formula_as_formula, true).c_str(), CE_UTF8) : NA_STRING;
	t++;
	}
	// Pad rows to even width
	if(contents[i].size() < out_width){
	unsigned int row_width = contents[i].size();
	for (unsigned int j = 0; j + row_width < out_width; j++){
	cell_values[t] = "";
	t++;
	}
	}
	}
	return cell_values;
	}