WIP: Fix FFI UB with arrays with child data

[ritchie46]

Which issue does this PR close?

See description in #20. I put this PR in a WIP state and need some help with this one.

[codecov-commenter]

Codecov Report

Merging #21 (d058927) into master (463f88f) will increase coverage by 0.02%.
The diff coverage is 93.33%.

❗ Current head d058927 differs from pull request most recent head e9874ad. Consider uploading reports for the commit e9874ad to get more accurate results

@@            Coverage Diff             @@
##           master      #21      +/-   ##
==========================================
+ Coverage   82.49%   82.51%   +0.02%     
==========================================
  Files         162      162              
  Lines       43522    43510      -12     
==========================================
- Hits        35902    35901       -1     
+ Misses       7620     7609      -11     

Impacted Files	Coverage Δ
arrow/src/ffi.rs	81.76% <87.50%> (+0.81%)	⬆️
arrow/src/array/ffi.rs	94.73% <100.00%> (+9.60%)	⬆️
arrow/src/array/transform/fixed_binary.rs	84.21% <0.00%> (+5.26%)	⬆️

---

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 463f88f...e9874ad. Read the comment docs.

[jorgecarleitao]

I am investigating this one.

[ritchie46]

I am investigating this one.

Great! I got stuck on this, so hopefully I can learn something from it.

[nevi-me]

I don't know if I'd be useful here, but I'll also have a look at this.

[ritchie46]

Close, fixed in #287