293 implement dust attempt2

[ThomasCeulemans]

Contains basic continuum/dust implementation; required input is grid of opacities at given frequencies.

Features:

• can now specify a dust/continuum opacity at each position, frequency; and specify the dust temperature at each position
• automated test (NOTE: without velocity gradient) for the expected intensity/optical depth given only dust present.
• freely choose the frequency grid to use for imaging using model.set_custom_spectral_discretization (using Pybind to get the frequency grid directly from python)

It seems that the dust implementation causes no significant performance regressions (tested on vanZadelhoff benchmarks), so no different compile options are needed.

This branch should fix issue #293.

[ThomasCeulemans]

Evidently, I still forgot to write any extra (not in c++ code) documentation. Ah well, this should be a reminder for me to actually do this and add an extra example to the documentation, or make an extra issue.

[ThomasCeulemans]

With the latest commit, issue #294 should be fixed.

[ThomasCeulemans]

Ah right, conda randomly decided to break CI procedures by adding tos confirmation on their command line interface... Well for unknown reasons the macos build still works fine, while the ubunu one hangs. Ah well, I'm confident that the linux build still works either way.