Develop fol

[loforest]

Description:

This PR solves a small bug, simplifies one calculation and adds two extensions to the algorithm:

• Bug: the LAYER_BASE_FRAC was not passed to the metarize method, so changing this parameter had no effect.
• Simplification: the parameter LAYER_BASE_FRAC was renamed LAYER_BASE_PERC, but has now a different behaviour. Instead of computing the layer height as the median of the lowest percentage of cloud heights, it is sopmuted as a given percentile. You obtain the same behaviour as before by setting LAYER_BASE_PERC = LAYER_BASE_FRAC*100/2
• Extension 1: calculation of cloud layer height as percentile of the most recent cloud hits instead of the whole time interval. This allows reducing the height underestimation of layers whose height is gradually increasing.
• Extension 2: merging of too close layers was introduced also in the grouping and slicing steps (not only the layering). In addition, the minimum separation parameters can be specified according to height ranges. This reduces the amount of algorithm overlayering, which was sometimes producing two FEW layers rather than one BKN layer (especially at high heights).

This PR is not yet finished: extension 2 does not work yet as specified and tests need to be added. Also, some tests did not pass, e.g. due to moving out from the layering step the AmpycloudWarning for too close layers.

Error(s) fixed:

Please tag them using #, e.g. Fixed #42.

Checklists:

• New code includes dedicated tests.
• New code has been linted.
• New code follows the project's style.
• New code is compatible with the 3-Clause BSD license.
• CHANGELOG has been updated.
• AUTHORS has been updated.
• Copyright years in module docstrings have been updated.

[fpavogt]

This PR has become obsolete in view of #86. I'll close it.