Properly support sliced unions

[jleibs]

I believe this resolves: #53

Rather than creating and maintaining parallel iterators, this uses the UnionArray::index() method to find the correct offset value to deserialize from and then deserializes from a newly created slice at that location in the correct child-array. One nice aspect to this approach is it generalizes to Sparse unions since index() does the correct book-keeping for us.

[codecov-commenter]

Codecov Report

Merging #91 (c447607) into main (dfa59fa) will decrease coverage by 3.88%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main      #91      +/-   ##
==========================================
- Coverage   98.25%   94.36%   -3.89%     
==========================================
  Files           8        8              
  Lines        1490     1544      +54     
==========================================
- Hits         1464     1457       -7     
- Misses         26       87      +61     

Impacted Files	Coverage Δ
arrow2_convert_derive/src/derive_enum.rs	98.96% <100.00%> (-0.07%)	⬇️
arrow2_convert_derive/src/derive_struct.rs	89.10% <0.00%> (-10.65%)	⬇️
arrow2_convert_derive/src/input.rs	80.00% <0.00%> (-6.51%)	⬇️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[ncpenke]

Thank you for the contribution and simplification @jleibs!

I was considering a similar approach, but wasn't sure what the performance implication of going through the full array deserialization would be for each union element. However, we don't have benchmarks around that yet, and this seems to be a reasonable trade-off for enabling more functionality, so will go ahead and merge this PR.

There's some minor code cleanup but I'll take care of that in another pass.