refactor: params extra types are zero values not nil pointers by default

[ARR4N]

Why this should be merged

Adds support for type-safe setting of ChainConfig and Rules extras via the ExtraPayloadGetter, which can now return pointers. This is demonstrated to work with zero-values of both types.

getter := params.RegisterExtras(Extras[Foo,...]{})

var x ChainConfig
fmt.Println(getter.FromChainConfig(&x)) // prints a zero Foo
*getter.PointerFromChainConfig(&x) = Foo{...} // returned pointer is guaranteed non-nil so this is always safe

How this works

I suspect that zero-value extras via x := new(ChainConfig) or var x ChainConfig have always worked this way, but the behaviour is now locked in with a test that also demonstrates shallow copying.

Note

As the extras are wrapped in a *pseudo.Type, shallow copying only copies the address of the wrapper even if the registered type is not a pointer.

Modification via pointers is achieved by extending the pseudo package to allow for "addressing" via the new PointerTo() function. This is then exposed via the ExtraPayloadGetter.PointerFromChainConfig() *C and PointerFromRules() *R methods to provide type safety to external users. The guarantees of correctness that avoid MustNewPointer() panicking are the same as those for the original FromChainConfig() and FromRules() methods.

The other changes to C and R plumbing (i.e never using *C and *R) were to clean up a code smell; this made the rest of this PR much easier to reason about. I relaxed the constraint on C and R being structs, allowing for pointers to structs, also because it's a smell (that I think can actually be removed entirely but in another PR).

How this was tested

New unit test for pseudo.PointerTo() and integration test for ExtraPaloadGetter.PointerFrom{ChainConfig,Rules}().