The generic HashTableBase approach was producing the wrong results for
the over-allocation, so I'm using the HashTable template now, which
seems to produce the right results.

Also distinguish properties backing stores for prototypes from regular
properties backing stores (since we're primarily interested in the
prototypes for now).

Bug: v8:7266
Change-Id: I5bbda6851f0320168ada1beb104042d0052c9a17
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1559869Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#60737}