Make serialization of slendr models to disk optional (for faster msprime coalescent runs)

[bodkan]

This is is an attempt to implement #97. Skipping serialization to disk makes it possible to call our msprime back-end Python code directly, without having to go through the standard Bunch of Files on Disk (TM) format of slendr models (normally needed to call the SLiM back-end script on the command line). This will, in turn, make coalescent simulations even faster, opening up the possibility to develop model fitting procedures in other projects (that functionality won't be part of slendr).

[bodkan]

Things are looking good so far. The only thing necessary was to make the back-end msprime simulation code into its own function. Then, compile_model was given a serialize = TRUE | FALSE argument (default TRUE), which allows skipping the writing of model configuration files to disk. The msprime() function can detect this and simply use R data frames (converted to pandas DataFrames) containing the model configuration data stored in memory (not on disk) and plug them into the msprime Python coalescent script.

More unit tests are needed to make sure I didn't miss some combination of missing gene-flow events, now resize events, etc. -- this would make some of those data.frames NULL/None, which needs to be take care of accordingly.

[codecov-commenter]

Codecov Report

Merging #112 (0ad6407) into main (b0bc2a0) will increase coverage by 0.21%. The diff coverage is 85.93%.

@@            Coverage Diff             @@
##             main     #112      +/-   ##
==========================================
+ Coverage   82.09%   82.31%   +0.21%     
==========================================
  Files           6        6              
  Lines        2905     2935      +30     
==========================================
+ Hits         2385     2416      +31     
+ Misses        520      519       -1     

Impacted Files	Coverage Δ
R/compilation.R	89.31% <85.71%> (+0.77%)	:arrow_up:
R/tree-sequences.R	87.22% <100.00%> (ø)

:mega: We’re building smart automated test selection to slash your CI/CD build times. Learn more