Plot the ribosomal subunit abundance and commit the results and figures

[edkerk]

The published analysis saved a density plot of the average abundance of the candidate
ribosomal subunits (average_riboSubunit_abundance.pdf). The Python version did not
write it. It is the only saved figure that was missing; the two enzyme-usage figures
already exist.

• python -m overflow.build_ribosome now also writes
  results/modelGeneration/average_riboSubunit_abundance.pdf and the numbers behind it,
  averageRiboSubunitAbundance.tsv.
• The plot is the published one: a kernel density (bandwidth 0.1 in log10) of the average
  abundance over every replicate of every condition, with MATLAB's default axes, fonts and
  title. A subunit missing from any replicate has no average, as in MATLAB's mean
  (157 of the 177 measured candidates), which also brings the range back to the
  published -7.3 to -3.9. The selected core is unchanged (48).
• candidate_means gives the averages for every measured candidate; core_subunits
  selects exactly as before.

Tests: the density integrates to one and peaks where the data do, the figure reports
the core count and is written, and on the real data the plotted means agree with the
selected core. 196 tests pass on Gurobi.

All three figures are drawn as the published ones: the usage figures follow
boxplotEnzymeUsage.R (theme_classic, 7 pt text, 10 x 4.5 cm, the published panel colours,
unfilled boxes with a doubled median line, solid outlier points) and the subunit figure is
MATLAB's default density plot. Helvetica is used where installed, else the metric-compatible
Nimbus Sans. They are included under results/ from a full run of the analysis (ecModels
gecko4 model, --fit-rates, proton leak closed with --block r_2129): enzymeUsage/selectedSystemUsage.pdf,
enzymeUsage/supplementSystemUsage.pdf and modelGeneration/average_riboSubunit_abundance.pdf.

The tables of that run are committed too, as the published repo did with its results: the
enzyme-usage tables, the model-generation tables, allFluxes_<cond>.tsv for each condition,
the sampling tables and COMPARISON.md, all under results/. The built models
(models/ecModel_P_<cond>.yml, about 67 MB) are not committed and rebuild in a few
minutes with the README recipe.