`ark-cli`: Server mode and text recognition

[Serve](https://github.com/ARK-Builders/ark-rust/issues/47) sub-command could be used for server-side text recognition. The product of this process would be text layers attached to graphical (and, maybe, audio) resources. Similar to previews, text layers could be stored somewhere in `.ark` folders, e.g. `.ark/text-layers/`.

Start by reading the linked Serve sub-command issue and then inspect how .ark folders currently store previews. Define how server-side text recognition would attach text layers to graphical or audio resources and where identifiers such as .ark/text-layers/ fit. Done should include an agreed design for generation and storage, which this issue does not yet specify.

Written by the indexing model from the issue text.