Improvements to figure 1

Draft figure from #57

What it definitely needs

[ ] Cartoon is a placeholder, I'm working on a replacement. UPDATE: now closer as of #63
[x] ~~Ticks from age histograms should line up with timeline in the cartoon~~
[x] ~~Mantel x and y labels are incorrect~~
[x] ~~PERMANOVA and Mantel for maternal education are not correct, need to transform -8 values to missing~~
[x] ~~PERMANOVAs and Mantels are currently based on the full dataset, which includes repeat measures. Need to stop that.~~
[x] ~~Need colorbars for ordinations and omnibus tests~~ I decided against this, there are already numbers in the cells
[x] ~~Need different colorscale for ordinations or for Mantel tests~~
[x] ~~Layout changes - possibly make it wider, make 1st column take more horizontal width, and put ordinations all on one row.~~
[x] Panel labels
[x] text sizes in permanovas, and in other places

Why does age have so little effect on ECs and KOs? Possible it's just because there's so little annotated at this level, it's hard to pick up on differences. Also possible that they are high-level annotations and more reflective of constant microbial niches.

ordinations are currently MDS1, MDS2, colored by age. This doesn't make much sense for metabolites especially. Should we do Age and MDS1 instead? If so, what should we color by? Alpha diversity?
Ombnibus tests are currently aligned by the text. Should we align by box instead? There are 5 columns in permanova, and only 4 in the mantel.
For ordinations, consider using transparancy
Consider also using lines to connect samples from the same kids.