Fix ldexp test errors

[devmotion]

This PR fixes the test error caused by JuliaDiff/DiffRules.jl#73:

julia>] test ForwardDiff
......auto-testing Base.ldexp with 2 arguments
ERROR: LoadError: MethodError: no method matching ldexp(::Int64, ::Float64)
Closest candidates are:ldexp(::ForwardDiff.Dual{Tx}, ::AbstractFloat) where Tx at ~/.julia/packages/ForwardDiff/tZ5o1/src/dual.jl:144ldexp(::ForwardDiff.Dual{Tx}, ::Real) where Tx at ~/.julia/packages/ForwardDiff/tZ5o1/src/dual.jl:144ldexp(::Integer, ::ForwardDiff.Dual{Ty}) where Ty at ~/.julia/packages/ForwardDiff/tZ5o1/src/dual.jl:145...
Stacktrace:
[1] ldexp(x::ForwardDiff.Dual{Main.DualTest.TestTag(), Int64, 1}, y::Float64)
@ ForwardDiff ~/.julia/packages/ForwardDiff/tZ5o1/src/dual.jl:243...

[devmotion]

Random values where x/y where (close to) an integer caused problems in DiffRules (JuliaDiff/DiffRules.jl#78), so I thought it's good to apply the same fix here as well.

[codecov-commenter]

Codecov Report

Merging #568 (6268c6c) into master (ab0e239) will increase coverage by 0.01%.
The diff coverage is n/a.

@@ Coverage Diff @@## master #568 +/- ##
==========================================
+ Coverage 85.36% 85.38% +0.01% 
==========================================
Files 9 9 Lines 868 869 +1 ==========================================
+ Hits 741 742 +1 
Misses 127 127 

Impacted Files	Coverage Δ
src/dual.jl	74.83% <0.00%> (+0.08%)	⬆️

---

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 ab0e239...6268c6c. Read the comment docs.