Next iteration of Namespace support

[aucampia]

We need to plan for the next iteration of namespace support. The current namespace support is a bit hard to work with as it stands, and we should try and simplify it.

This description will serve as somewhat of a summary of the discussion below and will be continuously updated, so please quote any part of it you refer to.

Some issues that should be addressed:

• rdflib 6.2.0 SPARQL parsing regression: multiple prefixes for same IRI not supported? #2077
• Incongruity in type signature of NamespaceManager #2004
• poorly documented behaviour of namespace manager #1584
• Improve Namespace attribute handling #1586
• Change bind() to be predictable #1616
• Non-deterministic prefix creation #1618

Relevant issues should match these labels:

• concept: namespace

Relevant discussions should be discoverable with this query:

• is:open label:"concept: Namespace"

Some questions that need answering:

• What should the data model be for namespaces and prefixes? Should it be Dict[Prefix,URI] or Sequence[Tuple[Prefix,URI]], or something else?
• What should be the constraints on prefixes and URIs be?
• How should users be able to interact with the prefixes?
• How should ambiguity be resolved in various cases like parsing and serialization?

Some concerns:

• There is coupling between NamespaceManager, Graph and Store - we should try and eliminate the coupling, but it will break the store interface, and we should see how we manage this best.