refactor benchmarks to avoid DeferredPromise

[staltz]

Two main changes here:

• Removed the DeferredPromise in benchmarks/index.js
  • I thought this was the only way of doing it, but I figured out a simpler way, which is just await new Promise((resolve) =>
• Moved the "add 1000 messages" benchmark to the bottom
  • I noticed that since this was the first test, then it meant that all subsequent tests were running over 101k messages, not 100k messages as ssb-fixtures was configured for. I guess we should stick to 100k messages for the tests because it helps us know how much time "per message" on average is spent. We should not add messages to these tests.

[github-actions]

Benchmark results

Part	Duration
Migrate (+db1)	9146ms
Migrate (alone)	4324ms
Migrate (+db1 +db2)	7737ms
Migrate (+db2)	3760ms
Migrate continuation (+db2)	1311ms
Memory usage without indexes	764.20 MB = 34.33 MB + etc
Initial indexing	1111ms
Initial indexing maxCpu=86	5865ms
Initial indexing compat	945ms
Two indexes updating concurrently	1216ms
key one initial	53ms
key two	3ms
key one again	6ms
reboot and key one again	69ms
latest root posts	593ms
latest posts	46ms
votes one initial	693ms
votes again	1ms
hasRoot	424ms
hasRoot again	1ms
author one posts	371ms
author two posts	30ms
dedicated author one posts	426ms
dedicated author one posts again	1ms
add 1000 elements	184ms
Maximum memory usage	791.80 MB = 43.39 MB + etc
Indexes folder size	12.23mb

[arj03]

I was looking at the old file and the new and this does seem to improve readability overall. This will probably mean I need to rebase some PRs but that is okay, I want to rebase them on top of validate2 PR anyway.

[staltz]

@arj03 if you want, I can do this refactor after the big merges happen.

[arj03]

Might be a good way to review it :)

[staltz]

I'll park this PR for now, I can copy-paste or reuse parts of it later.