Fix implementation of rand

[ablaom]

Closes #65.

This PR also changes the behaviour of rand(d, args...) to use default_rng in place of
GLOBAL_RNG, mimicking changes made to Random in Julia 1.7.

edit

No, rather we now address #65 by implementing Random.Sampler(rng, ::UnivariateDistribution) and Random.rand(rng, sampler) instead directly overloading each variant of rand([rng, ], ::UnivariateDistribution, args...). This has the effect of automatically using default_rng as the default rng, instead of GLOBAL_RNG, because that how Random overloads Base.rand.

[codecov]

Codecov Report

All modified lines are covered by tests ✅

Comparison is base (66f23a3) 95.85% compared to head (db73962) 96.03%.
Report is 3 commits behind head on dev.

Additional details and impacted files

@@            Coverage Diff             @@
##              dev      #66      +/-   ##
==========================================
+ Coverage   95.85%   96.03%   +0.18%     
==========================================
  Files           8        8              
  Lines         434      429       -5     
==========================================
- Hits          416      412       -4     
+ Misses         18       17       -1     

Files	Coverage Δ
src/methods.jl	91.26% <100.00%> (+0.43%)	⬆️

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ablaom]

@rfourquet Can you confirm that the change from GLOBAL_RNG to default_rng() in this PR is appropriate?

[ablaom]

@rikhuijzer Do you have time and interest in looking over this?

[rikhuijzer]

I would suggest to cleanup the symbol loading a bit with these changes:

diff --git a/test/methods.jl b/test/methods.jl
index 83215a2..30516fb 100644
--- a/test/methods.jl
+++ b/test/methods.jl
@@ -1,16 +1,17 @@
 module TestUnivariateFiniteMethods
 
-using Test
-using CategoricalDistributions
-using CategoricalArrays
+import CategoricalDistributions: classes, ERR_NAN_FOUND
 import Distributions
-using StableRNGs
 import Random
-rng = StableRNG(123)
+
+using CategoricalArrays
+using CategoricalDistributions
+using Random: default_rng
 using ScientificTypes
-import Random.default_rng
+using StableRNGs
+using Test
 
-import CategoricalDistributions: classes, ERR_NAN_FOUND
+rng = StableRNG(123)
 
 v = categorical(collect("asqfasqffqsaaaa"), ordered=true)
 V = categorical(collect("asqfasqffqsaaaa"))
@@ -314,7 +315,7 @@ if VERSION >= v"1.7"
         @test [rand(d) for i in 1:30] == samples
 
         Random.seed!(123)
-        samples = rand(Random.default_rng(), d, 3, 5)
+        samples = rand(default_rng(), d, 3, 5)
         Random.seed!(123)
         @test samples == rand(d, 3, 5)
     end

To apply these changes, you can copy them into a file somefile.patch and use

$ git apply mypatch.patch

[rikhuijzer]

Apart from that, LGTM