[ADD] Support for empirical Fisher in KFAC

[runame]

Resolves #46.

I have already added a 'type-2' option for the fisher_type argument. Not sure if I like the name since this terminology is not very common, but using 'exact' instead lead to confusion before.

[coveralls]

Pull Request Test Coverage Report for Build 6787848309

• 11 of 13 (84.62%) changed or added relevant lines in 1 file are covered.
• 40 unchanged lines in 2 files lost coverage.
• Overall coverage decreased (-5.2%) to 87.25%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
curvlinops/kfac.py	11	13	84.62%

Files with Coverage Reduction	New Missed Lines	%
curvlinops/examples/utils.py	6	41.67%
curvlinops/fisher.py	34	28.85%

Totals
Change from base Build 6697342688:	-5.2%
Covered Lines:	698
Relevant Lines:	800

---

💛 - Coveralls

[f-dangel]

Looks good to me modulo the comments on documentation. Remind me whether there is already an issue to implement Fisher type-2 (edit: yes there is, #49).