Enable automatic rerun of post-processing when things change

[ntalluri]

Currently, SPRAS does not automatically rerun post-processing steps if the logic or configuration of those steps changes. (The only time the post processing is rerun is if the algorithms change in some manner - parameters or adding/removing an algorithm). This can lead to outdated results being used without the user's awareness. This will be a problem for the continuous benchmarking.

We want to expect that if a user updates a post-processing function or its configuration, SPRAS should automatically mark dependent outputs as outdated and rerun the corresponding steps.

Proposed solution:

1. Add the config file as a parameter sent to each post processing step because it is timestamped and will trigger reruns.
2. Otherwise Add logic to detect changes in post-processing functions or configuration files.

[ntalluri]

There’s a specific test case to consider:

If a user initially runs the ML step with kernel_density set to false, the PCA output will not include KDE or kde_peak values. Later, if the user updates the config to enable kernel_density = true and tries to run the evaluation, the logic breaks. To fix this, the user would need to manually delete the PCA files and rerun the ML step with KDE enabled; an inconvenient and error-prone process. This should be improved so that changes in config parameters automatically trigger the necessary steps to rerun.

[agitter]

I've been working on this for LPCA specifically in #500. There, the Snakemake rules use the params keyword and values read from the config file to tell Snakemake to rerun the rule when those values change. This should work broadly.