Develop new benchmarks comparing the sticky-air and ALE (free-surface) approaches

[naliboff]

Following the discussion in this forum post, a number of ASPECT users have developed test cases that illustrate where the current free-surface implementation exhibits issues in non-cartesian geometries and deviate from equivalent models using a stick-air layer.

#6284 largely fixes these issues, and once it is merged it would be a good idea to develop benchmarks similar to those in the forum post.