update software and improve global epistasis convergence

[jbloom]

This pull request is designed to address #158. In limited testing it may have improved correlation between global epistasis fits and single mutant func scores, with the biggest difference being stricter convergence criteria and more iterations.

Also updates software versions and makes notebooks no longer read configuration YAMLs but instead have variables explicitly passed from papermill.

Specifically, here are all the changes:

• The func_effects_global_epistasis notebooks now make plots showing correlation between mutation effects inferred for mutations by the global epistasis model versus the average effects of these mutations among all single-mutant variants. This helps with evaluating questions like this about whether single mutant estimates are being affected by the global epistasis model.
• Increased the number of iterations and the stringency of the convergence criteria for the multidms global epistasis model fitting per the suggestions here.
• Update conda environment to latest versions of altair (5.4 -> 5.5), baltic (0.2.2 -> 0.3), biopython (1.83 -> 1.84), entrez-direct (16.2 -> 22.4), markdown (3.4 -> 3.6), matplotlib (3.8 -> 3.9). minimap2 (2.26 -> 2.28), plotnine (0.12 -> 0.14), python (3.11 -> 3.12), ruamel.yaml(0.17 -> 0.18),snakemake(8.20.6 -> 8.25.3),alignparse(0.6.2 -> 0.6.3), andmultidms(0.3.3 -> 0.4.0). Note that installingmultidmsdowngradesaltair` to 5.1.2 due to this issue.
• Made some internal changes to code to adjust to version update of ruamel.yaml, as new version drops round_trip_dump and safe_load.
• Some notebooks previously read input parameters from the YAML configuration file, which is non-ideal as it meant that snakemake was not explicitly tracking what was passed to them. Fixed this: now all parameters they use are passed via papermill parameterization.