[v2.3] Binary population test suite and reorganizing dev-tools

[sgossage]

This is a cherry picked version of PR#779 for v2.3.

If you go into dev-tools you can run

./run_test_suite.sh <branch-name>

to run the full test suite. This will run the binary tests and then the population tests right after. Output is saved to logs/<branch-name>/evolve_populations.log. Populations that are run as a result of the test suite show up in output/population_tests/<branch-name>/.

This adds a population synthesis test suite and performs 4 tests currently:

• BinaryPopulation saving binaries to RAM.
• BinaryPopulation saving to hdf5, plus check saved/loaded DataFrame is equivalent to binaries in RAM.
• BinaryPopulation saving in multiple batch hdf5 files (i.e., w/ optimize_ram = True).
• PopulationRunner with multiple metallicities¹. TODO: add a flag to turn on/OFF
• Test full pop synth pipeline via CLI. Right now, there's a small test for this that uses 3 metallicities, run locally. Running our full eight can cause OOM errors.

This also reorganizes dev-tools as follows:

dev-tools/
├── run_test_suite.sh
├── validate_binaries.sh
├── generate_baseline.sh
└── script_data/
    ├── inlists/
    ├── workdirs/
    ├── baselines/
    ├── logs/
    ├── output/
    │   ├── binary_star_tests/
    │   └── population_tests/
    └── src/
        ├── binaries_suite.py
        ├── binary_test_cases.py
        ├── compare_runs.py
        ├── formatting.py
        ├── popsynth_suite.py
        └── utils.py

where the .ini files are generated on-the-fly in inlists/ and test output is saved to the various output/ subdirectories.

Footnotes

1. ⚠️ The PopulationRunner test is implemented but it can break without appropriate system resources. (The RAM fix helps here: PR#818, so this branch and v2.3 should be safe to use it). ↩