about the filter: There is no right or wrong. They all have their right to exist, they were developed and published by different people at different times. In the end, they do similar things, reducing speckle to a certian degree. So it really depends on your application and what you want to extract from the later product. If you need point-like or linear features, you select smaller window sizes to preserve edges, but if you are working on larger areas (e.g. agricultural fields), you can apply a more aggressive filter to get homogenous areas. The best advice I can give is to try and compare.
Filtering before terrain correction should be the better choice, because most filters were designed for data in slant geometry.
You only need Beta0 if you want to apply radiometric terrain flattening afterwards. If not, you should apply the filter on Sigma0.
This might also clarify: Radiometric & Geometric Correction Workflow