[WIP] Init frame covariance

Depends on !135 (merged) .

Aims at Resolves #137 (closed).

@joanvallve I tried to somewhat follow your suggestion here regarding #137 (closed). I'd like your opinion as I'm not quite sure of what I'm doing xD.

...where should I look?

The state block marginal covariance is not the feature covariance as you implemented. The frame covariance is a propagation of the last KF covariance (with the corresponding Jacobians plus the measurement covariance). I think it will throw an eigen's size error most of the cases.

I think that it should be implemented in the corresponding processor. First, because it is very application-related. Secondly, because the propagation maths are (or can be) implemented there. In that case, a new emplaceFrame could be added in the problem's API providing also its marginal covariance.

@joanvallve yeah I already know that this PR is mostly garbage. We discussed it a while ago with @jsola and this might be more tricky than expected (we need to re-integrate the covariance at some points). It is still open to remind us that this issue should be tackled. Actually I think here is an issue card somewhere so that we could close this PR.

Since #145 (closed) #137 (closed) are still open, I close this MR

closed