Cache model blobs for faster serialization and thread-safety

[lantiga]

Address #122 and #302
Eventually caching could be deactivated if no RDB saving or AOF rewrite are needed.

[codecov]

Codecov Report

Merging #331 into master will decrease coverage by 0.31%.
The diff coverage is 67.30%.

@@            Coverage Diff             @@
##           master     #331      +/-   ##
==========================================
- Coverage   56.75%   56.44%   -0.32%     
==========================================
  Files          27       27              
  Lines        5050     5062      +12     
==========================================
- Hits         2866     2857       -9     
- Misses       2184     2205      +21     

Impacted Files	Coverage Δ
src/backends/tensorflow.c	68.66% <57.69%> (-2.03%)	⬇️
src/backends/torch.c	84.37% <62.50%> (-0.96%)	⬇️
src/backends/onnxruntime.c	68.01% <100.00%> (-0.43%)	⬇️
src/backends/tflite.c	69.15% <100.00%> (-1.12%)	⬇️
src/libtorch_c/torch_c.cpp	51.15% <0.00%> (-3.47%)	⬇️

---

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 6f36dab...614372e. Read the comment docs.