Define Random.Sampler

[devmotion]

This PR defines Random.Sampler for Sampleables.

Motivation:
Recently I updated ConsistencyResampling and switched to the Random API. The package now works with all distributions for which the Random interface is implemented. Unfortunately, I noticed that it is not possible to use Distributions without committing type piracy since Random.Sampler is not implemented (an example of this "fix" is shown in https://devmotion.github.io/ConsistencyResampling.jl/dev/supported/#Vectors-of-distributions).

Follow-up:
There are some questions that could lead to additional PRs:

• Is sampler still needed or should one switch to Random.Sampler completely and deprecate sampler?
• Should one commit to the Random interface more closely? E.g., Random uses rand!(rng, x, sampler) whereas Distributions supports rand!(rng, sampler, x). Since it is not relevant for my use case I did not touch it in this PR but, if desired, one could implement rand!(rng, x, sampler) as well and possibly deprecate rand!(rng, sampler, x).

[codecov-commenter]

Codecov Report

Merging #1314 (95e7274) into master (99b2b57) will increase coverage by 0.02%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #1314      +/-   ##
==========================================
+ Coverage   81.48%   81.51%   +0.02%     
==========================================
  Files         115      115              
  Lines        6623     6626       +3     
==========================================
+ Hits         5397     5401       +4     
+ Misses       1226     1225       -1     

Impacted Files	Coverage Δ
src/genericrand.jl	100.00% <100.00%> (ø)
src/univariate/discrete/poisson.jl	70.14% <0.00%> (+0.45%)	⬆️
src/matrix/matrixnormal.jl	94.11% <0.00%> (+1.47%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 99b2b57...95e7274. Read the comment docs.

[matbesancon]

All good here, I haven't looked at the API, but simplifying things in Distributions itself wouldn't hurt so if the Random.Sampler API is redundant we could only keep this one.
If we go in this direction, it needs to be done before 1.0. Could you add an issue to track this?

[devmotion]

I opened an issue: #1316

[matbesancon]

@devmotion we can merge this one too for the release

[devmotion]

OK, I fixed the merge conflict. I can merge and tag 0.25 if the tests pass?

[matbesancon]

yes absolutely