Added tests for gcd and lcm for other integer data types.

[ravibitsgoa]

Added tests for all signed and unsigned integer data types.
Improved the doctests of gcd and lcm.

Currently gcd(-5) = -5. Should we keep it that way or make it abs(-5) = 5?

[ravibitsgoa]

Can anyone review this PR which adds tests for previously untested integer data types for gcd and lcm?
Previous code tested gcd and lcm methods for only Int32 and Int64.

[KristofferC]

This made different packages start to error with a method error for checked_abs (e.g. https://github.com/JuliaCI/NanosoldierReports/blob/90885ffd4005a91d0dc4150b8c1453f2f5948eb0/pkgeval/by_hash/194ddda_vs_b5f13c3/logs/Polymake/1.5.0-DEV-f11a73bd9c.log). cc @StefanKarpinski

[StefanKarpinski]

This puts us between a rock and a hard place: on is it better to break compatibility or give the wrong answer sometimes if there's overflow? One solution would be to add a checked_abs(x::Number) = abs(x) fallback, but I think that's too dangerous. Another would be to used

checked_abs_if_available(x::Number) = applicable(checked_abs, x) ? checked_abs(x) : abs(x)

for this code in gcd an lcm. That would allow it to keep working for types that haven't defined checked_abs but not foist that unsafeness on anyone who explicitly calls checked_abs. It might be even better to put an off-by-default deprecation warning in the definition saying "please define checked_abs(::T)."