Use CidHashMap in the plain CAR-file blockstore

[jdjaustin]

Summary of changes

Changes introduced in this pull request:

Replace maps used for write cache and index in src/db/car/plain.rs with CidHashMap.

• Add missing methods to CidHashMap
• Replace HashMap<Cid, V> with CidHashMap<V> in car/plain.rs
• Measure change in peak memory use when validating an uncompressed calibnet and mainnet snapshot. See /usr/bin/time -v forest-cli snapshot validate {snapshot}.car

Reference issue to close (if applicable)

Closes #3286

Other information and links

Change checklist

• I have performed a self-review of my own code,
• I have made corresponding changes to the documentation,
• I have added tests that prove my fix is effective or that my feature works (if possible),
• I have made sure the CHANGELOG is up-to-date. All user-facing changes should be reflected in this document.

[jdjaustin]

Running gtime -v forest-cli snapshot validate forest_snapshot_calibnet_2023-08-01_height_786458.forest.car with the latest changes from this PR (peak RSS 593760 kbytes) versus main branch (peak RSS 855152 kbytes):

[jdjaustin]

Running gtime -v forest-cli snapshot validate 3095520_2023_08_04T18_00_00Z.car with the latest changes from this PR (peak RSS 6432096 kbytes) versus main branch (peak RSS 6473984 kbytes):

[lemmih]

Try running it with --check-stateroots 0.

[lemmih]

Better yet, use forest-tool benchmark graph-traversal

[lemmih]

I'm quite surprised by your numbers. They should be higher, and there should be a larger difference between your branch and main. These are the numbers I get:
main/graph-traverse: 13.26 GiB
main/validate: 17.74 GiB
josh/graph-traverse: 9.26 GiB
josh/validate: 13.75 GiB

I'll try running it on MacOS.

[jdjaustin]

My machine has done weird things ever since I got it. I recompiled this branch and ran it again against main; this time I'm getting much more expected results.

[jdjaustin]

I sense merge conflicts incoming due to #3388 merging soon, so converting this to draft until after I resolve those.

[ruseinov]

It would be nice to have some unit tests for the new CidHashMap functionality.