fix: replace fast-atomic-write with steno

[SgtPooki]

fixes #284

fast-write-atomic hasn't been updated in 5 years, is CJS, and is slower than steno (updated 2 months ago).

Benchmarks for various content-types & libraries (though we only use Uint8Arrays) can be found at https://github.com/SgtPooki/fast-write-atomic#benchmarks

However, there may be further room for improvement by moving to fs.createWriteStream

╰─ ✔ ❯ hyperfine --parameter-list branch 284-chore-replace-fast-write-atomic-with-steno,main --setup "git switch {branch} && npm run reset && npm i && npm run build" --runs 20  -w 1 "npm run test:node"
Benchmark 1: npm run test:node (branch = 284-chore-replace-fast-write-atomic-with-steno)
  Time (mean ± σ):     27.212 s ±  0.832 s    [User: 34.810 s, System: 6.051 s]
  Range (min … max):   25.927 s … 29.324 s    20 runs

Benchmark 2: npm run test:node (branch = main)
  Time (mean ± σ):     42.971 s ±  0.637 s    [User: 35.297 s, System: 7.534 s]
  Range (min … max):   42.178 s … 44.796 s    20 runs

Summary
  npm run test:node (branch = 284-chore-replace-fast-write-atomic-with-steno) ran
    1.58 ± 0.05 times faster than npm run test:node (branch = main)

---

Updated benchmarks of npm run test as of 2024-04-19

╭─    ~/code/work/protocol.ai/ipfs/js-stores    main ?1 
╰─ ✔ ❯ hyperfine --parameter-list branch main,test/not-same-event-loop-concurrency,284-chore-replace-fast-write-atomic-with-steno --setup "git switch {branch} && npm run reset && npm i && npm run build && cd packages/datastore-fs" "npm run test"
Benchmark 1: npm run test (branch = main)
  Time (mean ± σ):     99.415 s ±  2.918 s    [User: 69.659 s, System: 23.361 s]
  Range (min … max):   96.134 s … 105.200 s    10 runs

Benchmark 2: npm run test (branch = test/not-same-event-loop-concurrency)
  Time (mean ± σ):     103.456 s ±  3.186 s    [User: 74.442 s, System: 25.261 s]
  Range (min … max):   98.813 s … 108.429 s    10 runs

Benchmark 3: npm run test (branch = 284-chore-replace-fast-write-atomic-with-steno)
  Time (mean ± σ):     80.308 s ±  2.107 s    [User: 74.331 s, System: 22.228 s]
  Range (min … max):   78.219 s … 84.277 s    10 runs

Summary
  npm run test (branch = 284-chore-replace-fast-write-atomic-with-steno) ran
    1.24 ± 0.05 times faster than npm run test (branch = main)
    1.29 ± 0.05 times faster than npm run test (branch = test/not-same-event-loop-concurrency)                                                                                        [49m1.944s]

╭─    ~/code/work/protocol.ai/ipfs/js-stores    284-chore-re…c-with-steno ?1 
╰─ ✔ ❯ hyperfine --parameter-list branch main,test/not-same-event-loop-concurrency,284-chore-replace-fast-write-atomic-with-steno --setup "git switch {branch} && npm run reset && npm i && npm run build && cd packages/blockstore-fs" "npm run test"
Benchmark 1: npm run test (branch = main)
  Time (mean ± σ):     98.840 s ±  2.612 s    [User: 68.486 s, System: 22.585 s]
  Range (min … max):   97.005 s … 104.396 s    10 runs

Benchmark 2: npm run test (branch = test/not-same-event-loop-concurrency)
  Time (mean ± σ):     105.307 s ±  2.335 s    [User: 72.442 s, System: 24.766 s]
  Range (min … max):   101.167 s … 109.007 s    10 runs

Benchmark 3: npm run test (branch = 284-chore-replace-fast-write-atomic-with-steno)
  Time (mean ± σ):     77.012 s ±  1.829 s    [User: 74.442 s, System: 21.938 s]
  Range (min … max):   75.258 s … 80.825 s    10 runs

Summary
  npm run test (branch = 284-chore-replace-fast-write-atomic-with-steno) ran
    1.28 ± 0.05 times faster than npm run test (branch = main)
    1.37 ± 0.04 times faster than npm run test (branch = test/not-same-event-loop-concurrency)

[SgtPooki]

self review after removing cached writers. code change is much smaller now @achingbrain

[SgtPooki]

@achingbrain anything else needed for this one or can I merge it?

[achingbrain]

LGTM, just the question over the error detection we used to do for fast-write-atomic.