Changes made in dsp/fourier/internal/fftpack break the formatting used
there, so these are reverted. There will be complaints in CI.
[git-generate]
gofmt -w .
go generate gonum.org/v1/gonum/blas
go generate gonum.org/v1/gonum/blas/gonum
go generate gonum.org/v1/gonum/unit
go generate gonum.org/v1/gonum/unit/constant
go generate gonum.org/v1/gonum/graph/formats/dot
go generate gonum.org/v1/gonum/graph/formats/rdf
go generate gonum.org/v1/gonum/stat/card
git checkout -- dsp/fourier/internal/fftpack
The paper cited in the comment for the a > 1 (now a >= smallAlphaThresh) case mentions this approach as a small note on the final page, so there is no need to update the comments.
I also looked into numpy's logic for this case, but a) couldn't determine a source for it, except for what appears to be a more complex variant (Algorithm GS), and b) it's slightly slower than this method, at least for a >= 0.2.
benchstat:
```
name old time/op new time/op delta
GammaRand/case_0 89.4ns ± 5% 90.2ns ± 3% ~ (p=0.084 n=20+20)
GammaRand/case_1 212ns ± 3% 101ns ± 5% -52.28% (p=0.000 n=20+20)
GammaRand/case_2 25.4ns ± 3% 25.3ns ± 4% ~ (p=0.369 n=20+20)
GammaRand/case_3 48.6ns ± 6% 48.8ns ± 6% ~ (p=0.702 n=20+20)
GammaRand/case_4 47.1ns ± 6% 47.6ns ± 3% ~ (p=0.147 n=20+20)
```
stat/distuv: provide full test coverage for Gamma distribution and small fixes
* Removed redundant calculations in the small alpha version of Rand.
* Lowered the threshold for small alpha to 0.2 in accordance with
https://www4.stat.ncsu.edu/~rmartin/Codes/rgamss.R.
* Added the option to pass integration bounds to PDF normalisation check.
* Other small fixes.
