Tidy up the `instance`-based reasoning in `Data.Rational.Properties`

[jamesmckinna]

Working on #1711 / #2182 (and specifically: the knock-on changes arising from streamlining Data.Nat.DivMod) drew my attention to the very (overly!?) involved instance-based reasoning in Data.Rational.Properties concerning gcd and its role in the normalise function.

Suggest a careful weeding out of the accumulated cruft and verbosity there: the many examples of first proving a lemma to the effect that some quantity is n satisfies n ≢ 0, and then appealing to another lemma to turn that into an instance of NonZero n...

... a temporary, and tip-of-the-iceberg, case of which is in commit 6d819ae

[MatthewDaggitt]

That is indeed a nice clean-up!

[jamesmckinna]

But a note of caution: strive to avoid non-conservative/breaking changes in the style of #2230 !