[HalfSweet]

A drag along the ruler was taken for a click, and its segment dropped, whenever it covered less than 2 µs. Zoomed in to a few hundred nanoseconds per division the whole view is shorter than that, so no segment could be marked at all; short spans such as a fast SPI transfer, or a rise on an analog channel (where the edge measurement does not apply), could not be measured, listed in the data stream or exported.

• Tell a click from a drag on the ruler by how far the pointer moves, 3 px as on the tracks, and keep any segment that covers at least one sample. The segment changes only once the pointer has moved past those 3 px, so a click no longer flashes empty A and B flags.
• Read the segment's length on the ruler and in the status bar the way the edge measurement's Δt reads, exact to one sample with at most three decimals: one sample at 400 MHz now reads 2.5 ns rather than 3 ns, and 2.61 µs at 100 MHz no longer rounds to 2.6 µs. Lengths of a millisecond or more now show three decimals instead of four, as Δt does.
• Update DESIGN.md §6.6 and the user guide.

Not changed: the segment's start and end times in the status bar's edge table (four decimals of ms) and in the data stream's segment summary (three) keep their fixed precision, so the two ends of a very short segment can read the same there.

Validation:

• pnpm exec eslint src
• pnpm typecheck
• pnpm test: 573 tests passed, including the new TimelineRuler.test.ts and SegmentMeasure.test.ts, which fail on the previous code (a 500 ns drag at 100 MHz is dropped; one sample at 400 MHz reads 3 ns)
• Not checked by hand in the running app