Work on modifying initial proposal covariance for use with logit transform

[dmcdougall]

The logit transform:

1. isn't well documented;
2. is not in the user manual;
3. has complications that aren't explained;
4. has side-effects on printing the proposal covariance matrix which aren't documented;
5. ditto 4. but for writing to a file.

Also the initial proposal covariance is mapped to a different domain to do the sampling, and we should describe this procedure as well.