Introducing EncodedFSKeystore with base32 encoding

[AuHau]

Encoding the key's filename with base32 introduces coherent behavior
across different platforms and their case-sensitive/case-insensitive
file-systems. Moreover, it allows a wider character set to be used for the
name of the keys as the original restriction for special FS's characters
(e.g. '/', '.') will not apply.

Tightly coupled with migration: ipfs/fs-repo-migrations#84

[AuHau]

The failing test does not seem to be related to this PR.

Btw. I started to really like this feature, because it will enable to namespace the key's names like: websites/adam-uhlir.me

[Stebalien]

Thanks! @Kubuxu can you take a look at this? If it looks good, we can merge it after the release freeze (although we may want to coordinate with #5870 as that will also require a migration).

[Kubuxu]

Agree with after release, also I would wait with merging/migration until we can collect everything we might what to migrate.

[AuHau]

Fixed conflict.

Also one question, I have used mbase for the encoding/decoding instead of directly using base32. What do you think about it, is it alright? It sort of made sense to me...

[Stebalien]

For consistency, it might be better to just use base32. Using multibase doesn't buy us anything in this case.

[Kubuxu]

The only blocker is migrations at this point.

[AuHau]

Migration is almost ready at: ipfs/fs-repo-migrations#84
Only tests are missing, I will focus on them today and hopefully, all should be ready then.

[Kubuxu]

We will have other things in the migration: ipfs/fs-repo-migrations#85

[hsanjuan]

Continuing work at #6955

Thanks @AuHau for this!