[ENH] Write SNR0 in tSNR interface

[oesteban]

The TSNR interface should (optionally) write out the SNR0 map, since it is a required step for computing the tSNR.

Additionally, it should accept two masks to indicate where to measure the signal and where to measure the background noise.

https://doi.org/10.1002/mrm.1240

WDYT @chrisfilo ?

[cni-md]

Please keep in mind that this will only work with no / grappa acceleration and not with Sense. The roemer reconstruction gives you a homogeneous image intensity with an inhomogeneous noise distribution. In addition Philips will set voxel values <=5 to Zero -> no Backgroud noise. On a Philips scanner you have to perfom a dynamic noise scan with / without rf to calculate the SNR.
On a Siemens scanner the image intensity filter with a body coil ref image may also affect the noise distribution.

This function is nice to have but it has some pitfalls you should aware of.