We inline array allocation for wasm-gc in the TF graph by using
AllocateRaw nodes. Additionally, we use memset to initialize large,
zero-initialized arrays. These changes give measurable speedup in some
benchmarks.

Bug: v8:7748
Change-Id: Icbd37d0fe673c673379139b96d0e1c175e95e357
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3666618Reviewed-by: Maya Lekova <mslekova@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Commit-Queue: Manos Koukoutos <manoskouk@chromium.org>
Cr-Commit-Position: refs/heads/main@{#80762}