Add Downgrade CI

[fhagemann]

This was a bit of PITA and took quite some time, but it's for a good cause!

Downgrade CI can be used to check whether the versions bounds in compat are actually supported.

Why wouldn't that be? If a package introduces some functionality in its newest version and it is added to the project, the compat lower bound is not automatically updated because it might not be tested. Downgrade CI tests the package with the lowest allowed compat bounds and, then, runs the test. This way, compat lower bounds will also be tested and can be updated if necessary.

In SSD, three packages were listed with compat bounds too low:

• ProgressMeter: The ProgressMeter.ProgressThresh in convergence.jl#L22 uses keyword arguments that were introduced in ProgressMeter@1.5. In the Project.toml, versions down to ProgressMeter@1.2 were allowed.
• Unitful: The extensive use of uparse in ConstructiveSolidGeometry was only introduced in Unitful@1.0 and the support for range with units in MCEventsProcessing.jl#L149 for julia-1.8 (or newer) requires at least Unitful@1.11. In the Project.toml, the minimum version is still Unitful@0.18.
• IntervalSets: The function isclosedset used in ScalarPotential.jl#L124 was introduced in IntervalSets@0.5. The Project.toml allowed versions down to IntervalSets@0.3.

When running Downgrade, a series of other incompatibilities appeared:

• Some bug in StaticArrays@0.12 --> set lower bound to StaticArrays@1 --> requires lower bounds on Rotations@1 and Clustering@0.15
• Interpolations@0.14 --> requires lower bound on Requires@1.1 --> requires lower bound on ArraysOfArrays@0.5
• FillArrays@0.9 --> requires lower bound on Distributions@0.24.5
• LinearAlgebra, Random and Statistics require <0.0.1, 1 instead of 1 to support older julia versions.

While running the tests, some deprecation warnings for range(start, stop) appeared. They are also fixed in this PR.

[oschulz]

Very nice! Definitely something for the ToDo-list of lots of other packages ... :-)

[oschulz]

Please merge at your convenience, @fhagemann