Release allocated resources in Zlib, fixes #43.

[fredrikekre]

Together with @KristofferC I managed to find a solution to the memory leak issue (#43). It seems that the allocated resources in zlib are never released, but calling finalize does this.

It is a bit unclear from the TranscodingStreams documentation whether this is necessary, or if writing the END_TOKEN should do this. For example: https://juliaio.github.io/TranscodingStreams.jl/latest/#Wrapped-streams-1 hints that one must do something to release them, but in the code here we don't want to call close(zWriter) because that would also close the wrapped stream. https://juliaio.github.io/TranscodingStreams.jl/latest/examples/#Explicitly-finish-transcoding-by-writing-TOKEN_END-1 is a bit unclear what it means to "finishes the current transcoding process without closing the underlying stream", i.e. should that also release the allocated resources or not.

Draft for now, since it is possible this is a bug in TranscodingStreams (see JuliaIO/TranscodingStreams.jl#117)

[codecov-commenter]

Codecov Report

Merging #100 (1565685) into master (6ef44e4) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master     #100   +/-   ##
=======================================
  Coverage   96.01%   96.01%           
=======================================
  Files          16       16           
  Lines         803      804    +1     
=======================================
+ Hits          771      772    +1     
  Misses         32       32           

Impacted Files	Coverage Δ
src/write_data.jl	96.42% <100.00%> (+0.02%)	⬆️

---

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 6ef44e4...1565685. Read the comment docs.

[KristofferC]

Imo this should be merged since it fixes a real issue. If the API of TranscodungStreams changes, the code here can change too.

[jipolanco]

I agree with @KristofferC. I'll merge this if that's OK for @fredrikekre.

BTW thank you both for finding the solution!

[KristofferC]

Valgrind to the rescue :P

[fredrikekre]

Sounds good. I don't know what the failure on nightly is, but I don't think it is related.

[jipolanco]

Great, thanks! The failure on nightly doesn't seem to be related; it happens also on the current master.