We store deopt inline frames for all functions when we receive the code
creation event. We only ever use this information for code which is
deoptimized. Given that we receive code deopt events, we can just store
this information when the code is deoptimized.

At the time of the code deopt event, we also know the associated
deopt_id. That means we don't need to store a map of deopt_ids to
vectors of frames, because we will only ever access the frames for the
deopt_id that is already set.

This means we store way less data, particularly for long-running
processes which see fewer deopts. This saves 10MiB peak memory on the
node server example.

Bug: v8:7719
Change-Id: If6cf5ec413848e4c9f3c1e2106366ae2adae6fb1
Reviewed-on: https://chromium-review.googlesource.com/1050289
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Reviewed-by: Alexei Filippov <alph@chromium.org>
Cr-Commit-Position: refs/heads/master@{#53330}