Unreviewed, reverting 321448@main (c963a9e4dece)

https://bugs.webkit.org/show_bug.cgi?id=326412
rdar://188897444

This didn't quite work and could lead to a transient activation from evaluateJavaScript being
leaked. For instance:

1. The client calls [webView evaluateJavaScript:@"setTimeout(f, 10)" ...]. The forced user gesture
   gives the window activation at time T1. The pending timer keeps the gesture's token alive.
2. Before f runs, the client calls [webView evaluateJavaScript:@"setTimeout(g, 1000)" ...]. The
   second forced user gesture records T1 as the window's previous activation timestamp, and gives
   the window activation at time T2.
3. f runs and the first token is destroyed. The window's activation is T2, not T1, so nothing is
   revoked.
4. g runs and the second token is destroyed. The window's activation is T2, so it is restored to
   the previous activation that the second gesture recorded (T1).

This also caused a PLT regression.

Reverted change:

    Regression(316338@main) “Copy to clipboard” button on some sites sometimes does not copy the first time
    https://bugs.webkit.org/show_bug.cgi?id=324512
    rdar://186521201
    321448@main (c963a9e4dece)

Canonical link: https://commits.webkit.org/322775@main