Before this change large fixed arrays were visited in chunks of 32KB
on the main thread. Now large fixed arrays are divided into chunks
of kMaxRegularHeapObjectSize, which is consistent with large fixed
array processing on concurrent threads and make more sense considering
that regular fixed arrays are visited in one go.

Bug: chromium:1019218
Change-Id: I0c6d9c6c7d1db7f6a75d282601b73d4819b30e80
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1897886Reviewed-by: Dominik Inführ <dinfuehr@chromium.org>
Commit-Queue: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#64746}