radar: add in-situ averaging version

[rcknaus]

Implements a few averaging/filtering options.

Also fixes some get_if_present issues where I wasn't passing the default. Shouldn't affect things but it's better to be safe there.

Some details:

• I put some fixed number of points truncated-normal weighted Gauss quadratures in. The notation is truncated_normal{n} for the rule, where n refers to the number of sigma away the end of the cone is (which is where the distribution gets truncated). So higher n here refers to more heavily weighting points near the center of the spherical cap. 5 points are used for the quadrature but I could add more options there if there's a want for that. Alternatively, it might be better to just use the radau quadrature with a few built-in weight functions in the integrand.
• If the main axis is in the domain (or if always_output is on), I start reporting the average even if the full cone isn't in the domain. This means the average is spatially biased a bit until the cone is fully in the domain. Erring a bit on the side of over-reporting.
• The line-of-sight(site) component of the velocity is used along each quadrature ray, since the data wouldn't be available to do that in postprocessing any more.