Create a data storage library

Part of #554

Related to #818

Description

Goals:

• abstract source code of fetchers and services (Web API, indexation script, Python client) from storage specificities, letting them manipulate a domain-level data model instead of a storage-level one
• concentrate the source code handling data storage at one place
• improve the documentation of the storage model, with all its variants
  • git+tsv, git+jsonl, filesystem+tsv, filesystem+jsonl

Features:

• serialize and deserialize data model instances to many backend storages (e.g. git+tsv, git+jsonl)
• provider capabilities like accessing past revisions

Use cases:

• convert script of fetchers write data model instances to storage
• indexation script and web API read data model instances from storage
• simplify towards one storage model
• get rid of bare repositories by converting to git+jsonl the repositories that can't be checked-out because of the too many tsv files

Tasks

• create a new Python package: dbnomics-storage
• move and adapt parts of dbnomics_data_model.storage.* to new package
• adapt clients of dbnomics_data_model.storage to new package
• remove dbnomics_data_model.storage