WIP: start adding Transformer based audio encoder

[egpbos]

I just started a very rough first implementation, it's not working yet, but I'll just keep updating it in this work-in-progress PR.

[egpbos]

It works now (on my machine and on CI), suggestions / comments very welcome.

[egpbos]

Also implemented the optional config parameters scheme I suggested in #5.

[codecov-io]

Codecov Report

Merging #3 into master will increase coverage by 0.58%.
The diff coverage is 95.38%.

@@            Coverage Diff             @@
##           master       #3      +/-   ##
==========================================
+ Coverage   45.56%   46.15%   +0.58%     
==========================================
  Files          12       12              
  Lines         812      858      +46     
==========================================
+ Hits          370      396      +26     
- Misses        442      462      +20     

Impacted Files	Coverage Δ
platalea/encoders.py	26.81% <92.68%> (+3.15%)	⬆️
experiments/basic-stack-trafo/run.py	100.00% <100.00%> (ø)
platalea/basic.py	87.20% <0.00%> (ø)
experiments/basic-stack/run.py

---

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 70f50a3...259809b. Read the comment docs.

[bhigy]

Based on our recent discussion, I reverted the move of epochs from the experiment to the global config, as it doesn't make sense for some scripts (e.g. the utilities). We could create a function somewhere to add to the parser all experiment parameters that are generic like this one.

[egpbos]

Ok, thanks for the review! Applied your suggestions, merging now so I can train based on this model.