Move filesystem metadata root on-chain

[franciscoaguirre]

Currently, the filesystem metadata root exists only in the storage provider's memory. We can't verify proofs against it. Same as the MMR root, it should be on-chain so clients can request proofs that a file is at a particular path and validate them.

[bkontur]

If we move anything on-chain, how it would be used? For ensuring, we have challenges system. I mean, if the clients needs to constantly call RuntimeApi or storage items - it is against main principle:

The chain exists as a credible threat, not the hot path.

This is my main concern, if filesytem/s3 L1 implementations should be on-chain or purely off-chain provider implementation based on the bucket system - relates to: #51.