Fix SqExponential and GammaExponential + style

[willtebbutt]

Our implementations of the mentioned kernels were quite non-standard, so I've brought them inline with what is usually used.

I've also fixed some style issues. Apologies for conflating the two in a single PR.

The only changes of importance are those in src/exponential.jl -- everything else is either style or modifying tests to respect the new convention.

[willtebbutt]

Drop in coverage appears to be to do with style-related changes inflating the line count slightly.

[willtebbutt]

@theogf @sharanry the test failures appear to be unrelated to this PR. Any idea what's going on?

[theogf]

Looks like Gabor fails randomly...

[willtebbutt]

That's a bit sad. I'm reluctant to break the build, because that's really no fun for development. Have we got work ongoing to fix Gabor + AD?

[devmotion]

It seems to be fine apart from the tests with Julia nightlies, doesn't it? I assume the test errors there are caused by some Zygote and compiler-related issues that we do not have to worry about (at least I know that these Zygote/compiler problems cause test errors of Turing with Julia nightly quite consistently). Maybe best to just allow test failures on Julia nightly?

[willtebbutt]

Oh, weird. It was failing on version 1 before. I guess someone re-ran the builds? Assuming that docs pass then this is good to go.

[theogf]

Yeah I reran the failing build