This was achieved by exploiting the diagonality of the Fisher metric.
In cases this is not true one needs more samples to increase sensitivity.
Tested that a factor of 2 only in the inverse covariance can now be detected.

Fixed Fisher tests. The following bugs were present:

 - Metric of VariableCovarianceGaussianEnergy was wrong (already fixed in NIFTy+7, was fixed by merging)
 - debugging factor was not used, thus leading a false 'did not raise' assertion
 - the fast Fisher test for the VariableCovarianceGaussianEnergy was totally broken
 - slightly decreasing the tolerance for the isher test leads to factors of 2
   being detected

Log invgamma

See merge request ift/nifty!614

Various updates

See merge request ift/nifty!611

Implemente pre-nonlineariaties

See merge request ift/nifty!610

Fix slice operator

See merge request ift/nifty!606

Laplace distribution operator for L1 regularisation

See merge request ift/nifty!599

The information theoretic interpretation of drawing samples from the
metric of the KL samples is not clear. This may be the reason why it was
not documented before.

Update copyright year

See merge request ift/nifty!609

Allow the SliceOperator to handle a new shape which equals the old
shape.

Ducc is optional

See merge request ift/nifty!608

Add minisanity

See merge request ift/nifty!605

correlated_fields.py: Streamline Matern parameters

See merge request ift/nifty!601

Add finufft

See merge request ift/nifty!600

add clarification to einsum docs

See merge request ift/nifty!602

Adjust the slope of the Matern kernel such that it resembles the
corresponding parameter from the non-parametric amplitude model. Change
the docstring accordingly.

Furthermore, amend the docstring of the Matern kernel to state the prior
distributions (i.e. the moment-matching) for the various parameters.