Monitor history charts are unreadable with more than a handful of containers

[karoltheguy]

Overview

The CPU and Memory history charts do not scale past a small number of containers.

Details

Use Case / Steps:

1. Open the Monitor tab and select a server running 10 or more containers.
2. Try to identify which container is responsible for a spike.

Expected vs Actual:

• Expected: The chart makes the significant series identifiable at a glance.
• Actual: Most series flatline at 0 to 2% and occupy the same pixel row, and the legend consumes more vertical space than it earns.

Specific findings:

1. Both charts render an identical full legend, so 11 containers produce 22 legend swatches to convey 11 facts.
2. HISTORY_COLORS[i % HISTORY_COLORS.length] (static/main.js) cycles the palette, so past its length two containers share a line colour with no disambiguation.
3. Autoscaling to the maximum series flattens everything else into an indistinguishable band at the bottom.

Suggested fix:

Default to the top N series by current value, with the remainder collapsed into a muted "others" band, and render one shared legend above both charts rather than one per chart.

This overlaps #256 (isolate-on-click service selection for the CPU and Memory history charts) and should likely ship alongside it.