Replace HugrInternals::base_hugr with simpler methods

[aborgna-q]

HugrInternals::base_hugr returns the root Hugr used to implement a hugr view.
In practice, this call is used in a handful of cases:

• To access the metadata/optype for a single node.
• To access the ExtensionRegistry.
• To access the Hierarchy that linked to the HugrInternals::portgraph
• To run full-graph validation
• To render mermaid/dot of the full base graph
• In SiblingGraph, to avoid nesting views when creating one from a generic HugrView.

This works fine for our current definitions, but non-trivial wrappers like the one discussed in CQCL/tket2#778 are not necessarily based on a single base Hugr.

All but the last point I listed can be implemented as individual methods on HugrInternals instead,

• metadata(&self, Node) -> Option<NodeMetadataMap>
• optype(&self, None) -> Option<OpType>
• extension_registry(&self) -> ExtensionRegistry
• hierarchy would return a Cow<'_, Hierarchy> so we can either return a reference to the internally stored one, or compute one on-the-fly (for the CircuitHistory usecase).
• Rendering should use the existing portgraph directly.
• SiblingGraph shouldn't need to do the unwrapping. For non-dyn types all the calls through wrapping layers will hopefully be inlined and optimized away (we should benchmark this).

One drawback to keep in mind is that we no longer get the generic "escape hatch" that lets us recover the outer hugr from a filtered view. This means for example that focusing on a region will hide away non-local nodes outside the parent's hierarchy. I'd argue that those cases should use region-pointers instead. See #1642 #1554.