WeakRefStringArray: make eltype respect Missing

[alyst]

Enhances #17 so that WeakRefStringArray allows missing iff
the underlying WeakRefString Array allows missing.

Should fix the eltypes of the dataframes returned by CSV.jl.

Needs tests.

[codecov]

Codecov Report

Merging #20 into master will decrease coverage by 0.46%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master      #20      +/-   ##
==========================================
- Coverage   78.72%   78.26%   -0.47%     
==========================================
  Files           1        1              
  Lines          47       46       -1     
==========================================
- Hits           37       36       -1     
  Misses         10       10

Impacted Files	Coverage Δ
src/WeakRefStrings.jl	78.26% <100%> (-0.47%)	⬇️

---

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 69d5ff8...57aff55. Read the comment docs.

[alyst]

I've tried to fix 0.7 adding ncodeunits(), but there's another error when calling WeakRefStringArray(data::Vector{UInt8}, A::Array{Union{WeakRefString{UInt8}, Missing}, N}) ctor, which I don't understand.
Looks like the ctor doesn't recognize it should support Missing.
This doesn't happen on 0.6 and I don't have a working 0.7 setup.

[quinnj]

Closing in favor of #21 (hope you don't mind that I stole several bits from here though 😃 )

[alyst]

@quinnj Sure, no problem

[nalimilan]

As noted on #21, I think we still need this.