It seems that allocating smaller pages is actually quite a bit faster than
larger pages, probably because they can be cached by malloc. Let's see what the
bots say.

In a follow-up I'll check whether the segment-pool is actually beneficial or
whether we should just remove it.

This also drops SegmentSize::kLarge as a way to make compilation deterministic.
Turns out that by now we need >8mb anyway, and the previous 1mb wasn't enough.
At the same time the compiler was fixed to not rely on virtual addresses of
zone objects anymore, and there's a bot checking whether the snapshot is
determistic.

Change-Id: I38cbb0d209d68b3671fd38763b42714811f4223e
Reviewed-on: https://chromium-review.googlesource.com/c/1346370Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Commit-Queue: Toon Verwaest <verwaest@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57683}