defStateType should define `getPointRepr` instead

[dehann]

Towards Riemannian, so better to use getPointRepr over getPointIdentity?

DistributedFactorGraphs.jl/src/entities/State.jl

Line 344 in e7252bf

DFG.getPointIdentity(::Type{$structname}) = $point_identity

[Affie]

@defStateType already defines getPointType, originally I only wanted getPointType, but it looks like you were right to want getPointIdentity as well. It is quite handy and allows for constant propagation.
I see 2 options:

• Create 2 macros; @defManifoldStateType and @defLieGroupStateType. (my preference for a easier developer/user experience)
• Use kwarg like expressions in the macro, ie. identity=...

The preferred way to get a random initialization point on manifold is rand(M) and not to hard code some arbitrary constant.