Let a read be seen on a white page, and stop a narrow column claiming a label

[Paururo]

Two things I knew were outstanding, both verified by running before being fixed.

A read body nobody can see

ReadColoring::Uniform fell back to #b3bcc6, a grey written into the code
rather than taken from the theme. Measured:

	on the white page	on the dark page
#b3bcc6, as it was	1.92:1	9.27:1
a wash of the theme's own ink	3.21:1	4.29:1

One colour for both themes was wrong in both directions: a pileup a reader has
to hunt for in the theme that is the default, and one shouting in the other. It
is mix(theme.background, theme.muted, 0.65) now, which is what a large filled
shape needs and follows a theme a caller brought rather than only the two shipped
here.

I had this recorded the other way round, as "the pileup ignores --theme dark".
It does not: the background and the labels follow the theme and always did. What
ignored it was the read body, and the theme it failed in was the light one.

No committed figure changes. I checked that by rendering all twenty five
examples the way CI does, not by rendering the gallery, which is what I had been
doing earlier in this session and is not the same thing.

Fourteen hundred labels through each other

A position label is stood on end, so what it needs across its column is its
height rather than its length. Below that they overlap, and they were drawn
anyway. On thirty samples over fourteen hundred sites the figure carried 1,465
rotated labels and their tooltips; it carries 66 now, and drops from 4.30 MB to
4.03 MB. It is the same threshold the letters in the cells already use.

That change had been sitting on fix/snp-panel-density since before the rest of
this work, and the branch had gone stale around it: its diff against today's main
deletes 8,872 lines, including every test added since. The change itself is
twelve lines, so it is here rather than rebased, and the branch is gone.

And one I looked at and left

The ideogram's centromere is also a colour written into the code, #b03a2e,
which measures 6.02 on white and 2.96 on the dark page. That is under the 3:1
a filled mark wants, but only just, and inside my own rounding. Moving it to
#c04a3c clears both at 4.90 and 3.64 and changes one committed figure. Changing
what a committed figure looks like for four hundredths of a contrast point is not
mine to decide, so it is written down here rather than done.

Checks

Both guarded by asking the question rather than pinning the answer: one computes
the contrast of whatever colour the theme gives a read and fails below 3:1, which
the old grey does; the other counts the rotated labels a four hundred column
panel draws and fails at 400, which is what it drew before.