feat: rewrite dag stat to be awesomerer

[Jorropo]

Fixes #8794
Fixes #8791

Todo:

• Add tests!

This adds:

• skip-raw-leaves option, that avoids downloading raw leaves when you
  already know their size in the parrent block
• multithreaded download
• properly count blocks that show up multiple times with different
  multicodec
• do not count size of inlined blocks

Speed:
Way way faster (it's hard to make an objective number because that depends a lot on how much multithreading helps you and how many raw-leaves block you skip), in my case: >100x faster.

[BigLep]

@aschmahmann will provide first round of feedback to determine if this is something that we want to land.

[BigLep]

General comment: it looks like this PR is covering multiple issues. We think it's best to have the PRs separate in case there are different levels of alignment on the issues themselves.

[aschmahmann]

skip-raw-leaves option, that avoids downloading raw leaves when you already know their size in the parent block

This seems specific to dag-pb and won't work for anything else. I don't think we should be adding anything dag-pb specific in here. Almost all dag-pb data is UnixFS and so people could look at ipfs files commands for this.

multithreaded download

This seems good, although this will become problematic with the introduction of selectors as there is currently no parallelizable selector traversal. I'd rather invest time/code complexity budget in parallelizable selector traversals if you're interested.

properly count blocks that show up multiple times with different multicodec

Seems reasonable.

do not count size of inlined blocks

Seems reasonable.

[Jorropo]

@aschmahmann

Almost all dag-pb data is UnixFS and so people could look at ipfs files commands for this.

UnixFS and ipfs files by extension does not provide as good information:

• it counts identical blocks that show up multiple times, multiple times
• it counts inlined blocks
• it doesn't provide you the block count (btw if you only care about the block count and not the size, this code would also work with other codecs, I do say would because there is no block count only option and the usage of it is likely pretty limited)

[Jorropo]

Closed in favor of #8843, I do not plan to work on rawleaf skipping or multithreaded downloads anytime soon.