feat(types): make TernaryHV parametric over integer type

[cvigilv]

• Refactor TernaryHV to be parametric over integer type (T <: Integer)
• Add inner and outer constructors for type inference and conversion
• Provide type-specific constructors for TernaryHV{T}
• Update helper methods (copy, similar, normalize) to preserve type parameter
• Update tests to cover TernaryHV with various integer types

[cvigilv]

Mini benchmark:

julia> for T in (Int8, Int16, Int32, Int64)
           println("Benchmarking with $T")
           H = [TernaryHV{T}() for _ in 1:10]
           @show @benchmark sum($H)
           @show @benchmark prod($H)
           println()
       end
Benchmarking with Int8
#= REPL[7]:4 =# @benchmark(sum($(Expr(:$, :H)))) = Trial(4.655 μs)
#= REPL[7]:5 =# @benchmark(prod($(Expr(:$, :H)))) = Trial(2.729 μs)

Benchmarking with Int16
#= REPL[7]:4 =# @benchmark(sum($(Expr(:$, :H)))) = Trial(8.541 μs)
#= REPL[7]:5 =# @benchmark(prod($(Expr(:$, :H)))) = Trial(5.479 μs)

Benchmarking with Int32
#= REPL[7]:4 =# @benchmark(sum($(Expr(:$, :H)))) = Trial(17.209 μs)
#= REPL[7]:5 =# @benchmark(prod($(Expr(:$, :H)))) = Trial(11.667 μs)

Benchmarking with Int64
#= REPL[7]:4 =# @benchmark(sum($(Expr(:$, :H)))) = Trial(36.667 μs)
#= REPL[7]:5 =# @benchmark(prod($(Expr(:$, :H)))) = Trial(27.541 μs)

This is useful for optimizing bigger experiments.

[cvigilv]

Immediately merging this as I think it is a trivial addition.