test: a five second wait is tight under four workers

[xAlcahest]

Summary

Five batteries in a row on this machine each dropped one spec, a different one every time, and all of them on a five second wait: transport-after-open, silent-machine, video-aspect, cue-split. Every one passes alone.

The reason is written in e2e/README.md and the README had not noticed it: that section reasons about a battery that "runs one spec at a time", which stopped being true at N24. Four workers share the machine, so four apps are slower than one was, and five seconds that was generous for one is tight for four. A wait is not an assertion: it is how long the harness allows the app, and a check whose condition never becomes true still fails, only later.

Thirteen waits go to fifteen seconds, raised at their own call sites, which is what that section already says to do rather than moving the global multiplier.

Changes

• e2e/specs: thirteen timeout: 5000 waits become timeout: 15000, in current-line-bands, dividers, grid-columns, interface-scale, transport-after-open, waveform-sash and waveform-zoom.
• e2e/README.md: the paragraph that described a one-spec-at-a-time battery now says four workers, with the measurement that bought the change and the two waits deliberately left short.

What was left alone, and why

asr.spec.js looks for a dialog that may not be there at all and catches the timeout, so its five seconds is a cost ceiling and raising it would add ten seconds to every call that finds nothing. video-empty.spec.js asks a surface to stay unmapped rather than to become it, which is a different shape of wait.

Verified on Linux. Full gate green, battery 87 spec files and 514 tests in 5:50, and the gate's own verdict line reads GATE GREEN.