[NSETM-2279] Use shared memory to write partial DataFrames of features

[GianlucaFicarelli]

• Improve performance of report extraction and features calculation by writing partial DataFrames to the shared memory (or temp directory, if shared memory is not available).
  Both the used memory and the execution time should be lower than before, when processing large DataFrames.
• Use zstd compression instead of snappy when writing parquet files.
• When repo is pickled, extract the DataFrames only if they aren't already stored in the cache.
• Remove fastparquet extra dependency.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 93.70079% with 8 lines in your changes are missing coverage. Please review.

Project coverage is 91.49%. Comparing base (5602fd3) to head (7edbf63).

Files	Patch %	Lines
src/blueetl/repository.py	55.55%	1 Missing and 3 partials ⚠️
src/blueetl/store/parquet.py	90.32%	0 Missing and 3 partials ⚠️
src/blueetl/store/base.py	66.66%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #33      +/-   ##
==========================================
+ Coverage   91.47%   91.49%   +0.02%     
==========================================
  Files          45       45              
  Lines        2638     2703      +65     
  Branches      539      554      +15     
==========================================
+ Hits         2413     2473      +60     
- Misses        140      141       +1     
- Partials       85       89       +4     

Flag	Coverage Δ
pytest	91.49% <93.70%> (+0.02%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.