automatic DFT decimation for band-limited sources

[stevengj]

Now that we have the decimation option (#1678) and it seems to be working well, it would be good to make this the default. In particular, whenever we have a known bandlimited source (e.g. a gaussian, where we set the bandwidth as e.g. a tail of 1e-7), we can apply the formula from that issue to set a default decimation factor for any DFT accumulation.

For an API, I'm thinking of a default decimation factor of decimation=0, which will mean "use the automatic decimation if possible, otherwise set decimation=1". That way, the user can still override the DFT decimation by setting it to a positive value.