core: Run scripts registered at frame script phase in a seperate pass

[nivkner]

fixes #10443, and fixes #14629 for the version that does not include buttons.
when buttons are added to the dynamically constructed object, that is a separate issue of the frame script of the child happening before the parent's constructor finishes

im not sure i nailed down all the subtleties regarding the interaction between the cleanup pass and goto,
but the PR passes all the current tests on my machine, and the tests i made,
so i believe it is ready for review as is.

[nivkner]

additional tests, attached here:
frame_script_test1.zip
frame_script_test2.zip
suggests that rather than inserting to the queue and then ignoring the script while traversing the tree normally,
there might actually be a deferred registration.
so if there is already a script in place, it gets executed during the traversal,
and the new one is executed during cleanup.

[nivkner]

this test
frame_script_test3.zip
complicates matters, because while the script on the already registered frame does occur during regular traversal,
as well as the cleanup, the new script is executed.
in other words, the scripts are modified in place.

previous tests shown that scripts introduced during cleanup are not executed during traversal, when there were no scripts before. so next guess is that scripts registered during frame-script phase are enqueued for execution separately.

[nivkner]

is there a scenario where the queued_script_frame field of MovieClip, when it is populated, is different from current_frame during frame script execution?
i replaced all usage of it with current_frame, and all the tests pass.

my current_plan is to replace it with has_pending_script boolean, which depend on the existence of a frame_script in the current frame, and can be toggled by the cleanup pass.

EDIT: never mind. will keep queued_script_frame.

[nivkner]

fixed the behavior for the new test cases.

separated queued_script_frame into a boolean and a number, but no information should be lost (any case where the flag would be false, the script would skip, regardless of the value of the Option).

there are 2 cases of frame script behavior that i haven't resolved.
one is the aforementioned dynamic construction with buttons,
and the other is
frame_script_test4.zip
where, when using flash player, the frame script of Spawn seems to just disappear from the second cycle onward, (it doesn't reach the call to stop()).

since both are sufficiently different from the issue addressed by this PR, i will leave them for another time.

[Lord-McSweeney]

#12942 might be fixed by this.

[nivkner]

the PR does not fix #12942, confirmed by running against commit 2f8f340

[Molisson]

Assuming we're fine with this PR with no objections, it is well detailed, good job.
I ignored test files.