Develop image-based autofocus for LF arm

[ieivanov]

Similar to #19

We need to evaluate if O3 defocus is significant problem in the LF arm as well and implement a solution similar to #48. Here the mirror piezo stage has a larger travel range than we currently use and we may be able to shift the z-scan range to compensate for O3 defocus rather than moving the coarser KIM101 stage. This will get around problems with the non-encoded nature of the KIM101. On the other hand, implementing the same approach as in the LF arm could be more robust solution.