neatogen: do intermediate 'sep', 'esep' calculations on 'double'
Point positions were being stored as float
s within the expand_t
struct, but
then primarily used in computation with double
s. This meant unnecessary loss
of precision was occurring. This change is arguably a fix, as sep
and esep
are documented to have double
type¹ so users would reasonably expect
calculations on them to happen in double
space. This change squashes 8
-Wfloat-conversion/-Wconversion warnings.