Performance: Cache column indexes

[bocca]

Classes such as ColumnNameReader and ColumnNameMatchingValueReader compute column indexes for every row. This is not necessary and degrades performance, as indexes do not depend on the row index. The API could be refactored to allow to compute indexes only once at Header creation, and then use indexes only.

For example:

public interface ISingleCellValueReaderFactory
{
    ISingleCellValueReader TryGetReader(ExcelSheet sheet, IExcelDataReader reader);
}

public class ColumnNameValueReaderFactory : ISingleCellValueReaderFactory
{
    public string ColumnName { get; }

    public ColumnNameValueReaderFactory(string columnName)
    {
        ColumnName = columnName;
    }

    public ISingleCellValueReader? TryGetReader(ExcelSheet sheet, IExcelDataReader reader)
    {
        return sheet.Heading.TryGetColumnIndex(ColumnName, out int index)
            ? new ColumnIndexValueReader(index) : null;
    }
}

And then:

propertyMap.WithReaderFactory(new ColumnNameReaderFactory())