Fix BigFloat support by generalizing procf type constraint

[ChrisRackauckas-Claude]

Summary

• Fixed procf function to accept any Real type for the s parameter instead of Float64
• The previous s::Float64 type constraint caused a MethodError when using BigFloat with λ >= 6 (which triggers the ad_rand algorithm)
• Added tests for BigFloat support to prevent regression

Testing

Tested interface compatibility with:

• BigFloat: Now works correctly for all λ values (both count_rand for λ < 6 and ad_rand for λ >= 6)
• JLArrays: Broadcasting works correctly - pois_rand.(JLArray([5.0, 10.0, 15.0])) returns a JLArray{Int64, 1}
• Float32, Float16, Rational, Int: All work correctly
• JET type stability analysis passes for both Float64 and BigFloat

Changes

1. src/PoissonRandom.jl:85: Changed s::Float64 to s::Real in procf function signature
2. test/runtests.jl: Added @testset "BigFloat support" with tests for:
   • count_rand with BigFloat (λ < 6)
   • ad_rand with BigFloat (λ >= 6)
   • Statistical validity with BigFloat

All existing tests continue to pass.

cc @ChrisRackauckas

🤖 Generated with Claude Code