TODO: Consider adopting IPIP-499 default "modern" UnixFS CID profile

[rvagg]

Ref: ipfs/specs#499

We currently differ from the proposed unixfs-v1-2025 profile in the following ways:

• Directory wrapping: we wrap in a directory so we can put a name on the file, this has come up before as something we could potentially not do
• Hidden files: we include them
• HAMT threshold: Helia only supports what they're calling links-bytes

The benefit of conforming to a standard profile will be that when we pack a file or directory for a user, they end up with the same root CID as they get when they use any other conforming implementation. This may be of interest to the kind of people who care about NFTs or other items where reproducibility matters. It's also a general UX thing because people just don't tend to understand how all this works and complain or question when they end up with different CIDs, so consistency makes a class set of user complaints go away. How much do we care though?

[rvagg]

According to discussion in the issue the plan is to update Helia to support the proposed HAMT threshold, so we could choose to do best-effort now and match unixfs-v1-2025 when possible.

My only hold-up here is the wrapped directory we insert, we need to discuss that and consider whether it's important or not. If we conform otherwise then the second CID in the DAG will match a unixfs-v1-2025 one so there's that at least.