Improved read of metadata

[jorgecarleitao]

This PR simplifies the (sync) reading of metadata by avoiding pushing thrift decoding to the underlying reader, instead decoding directly from a temp buffer.

This is generally a tradeoff (memory usage vs CPU usage), but it is generally recommended to not deserialize stuff directly from the IO if there is not a strong reason to - we already do this in the async version.

[codecov-commenter]

Codecov Report

Merging #143 (9427962) into main (3774868) will increase coverage by 0.03%.
The diff coverage is 63.63%.

@@            Coverage Diff             @@
##             main     #143      +/-   ##
==========================================
+ Coverage   74.41%   74.44%   +0.03%     
==========================================
  Files          78       78              
  Lines        3603     3604       +1     
==========================================
+ Hits         2681     2683       +2     
+ Misses        922      921       -1     

Impacted Files	Coverage Δ
src/read/metadata.rs	65.85% <63.63%> (-0.82%)	⬇️
src/read/stream.rs	63.41% <63.63%> (+3.89%)	⬆️

---

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 3774868...9427962. Read the comment docs.