> Infer a metadata table from the filename. Table should look like this: group, concentration_um.

* read: data/raw/drug_A-J_1mM_500uM_200uM_100uM_50u...csv
    96 rows - groups: A B C D E F G H I J VEH BLANK

The name carries the doses in order, right after the group range: A-J, then ten
concentrations from 1mM down to 1uM. Ten labels, ten numbers - so A is 1mM and J
is 1uM, and the ninth group is 2uM.

VEH and BLANK come after the doses rather than among them, and there are no numbers
left for them. They are labels, not concentrations.

* write: metadata
    group  concentration_um
    A               1000.0
    B                500.0
    ...                 ...
    J                  1.0
    VEH                  -
    BLANK                -

Twelve rows. The two controls have an empty concentration rather than a zero,
because they were not dosed with nothing - they were not dosed. A zero would put
two invented points on the curve.
