Please treat the five uploaded CSV files as a single relational dataset and generate a professional, light-themed manufacturing dashboard containing four specific, visually distinct charts. I need you to ensure zero label overlapping by applying data filtering where necessary. First, create a clustered bar chart by joining the production and machine tables to compare the total 'TargetOutput' versus 'ActualOutput', grouped by 'LineID'. Second, design a Sunburst chart or nested Donut chart to visualize defects, placing 'DefectReason' on the inner ring and 'Severity' on the outer ring; crucially, filter this to display only the top 5 most frequent defect reasons to ensure the chart remains clean and readable. Third, build a dual-axis combination chart with 'Hour' on the x-axis, showing average 'PowerConsumption_KWh' as bars on the left axis and average 'Temperature_C' as a smooth line on the right axis. Finally, you must generate a matrix heatmap to diagnose waste, where the x-axis is 'Hour', the y-axis is 'MachineName', and the color intensity represents the Scrap Rate (calculated as ScrapParts divided by ActualOutput), but strictly filter the y-axis to show only the top 10 machines with the highest total scrap rates to prevent visual clutter.