random: remove useless @inline

[rfourquet]

I did more or less what was suggested here for files in "random/", i.e. remove @inline annotations and check if performance is degraded. According to "BaseBenchmarks/src/random", there are two regressions:

1. randn, for which I guess it's too critical, so I re-added @inline
2. randstring: by maybe about 30%. It was not obvious which set of @inline are contributing to the regression, but I would believe that it's not a time critical function, so that it would be OK. (EDIT: not anymore on master, see post below)

I didn't run the whole BaseBenchmarks suite locally, so will run it here to check for other possible regressions: @nanosoldier runbenchmarks(ALL, vs=":master")

[nanosoldier]

Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @ararslan

[rfourquet]

Interesting, I had performed the benchmark tests about 2 weeks ago, and in the meatime, the results changed: randstring is now not a regression anymore, and the new regressions, for randcycle, randperm and shuffle, are easily fixable by re-adding one @inline annotation. Will be easier to sell this PR :)
@nanosoldier runbenchmarks(ALL, vs=":master")

[nanosoldier]

Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @ararslan

[timholy]

LGTM. Nice work!