perf: fs optimizations - part 1

[littledivy]

• Flatten op arguments.

  As a result, truncateSync becomes 2.3x faster.

  # main
  () => truncateSync("test", 0)
  time 1029 ms rate 97181
  time 1013 ms rate 98716
  time 745 ms rate 134228
  
  # This patch
  () => truncateSync("test", 0)
  time 469 ms rate 213219
  time 447 ms rate 223713
  

• ~8x faster way to serialize op_stat_sync result to JS.

  # main
  () => lstatSync("test")
  time 767 ms rate 130378
  time 767 ms rate 130378
  time 745 ms rate 134228
  
  # This patch
  () => lstatSync("test")
  time 99 ms rate 1010101
  time 99 ms rate 1010101
  time 98 ms rate 1020408
  

• ~14x faster file copy on MacOS using APFS' clonefile

  # main
  () => copyFileSync("test", "test2")
  time 8646 ms rate 1156
  time 7578 ms rate 1319
  
  # This patch
  () => copyFileSync("test", "test2")
  time 738 ms rate 13550
  time 647 ms rate 15455
  

• Fast path for perms.read.check / perms.write.check to avoid resolve_from_cwd when possible. It's quite literally a bool check when run with -A, making it infintely faster than before.

[vimirage]

The struct writer should accept a Uint32Array -> &mut [u32]

[bartlomieju]

LGTM