and move to python3

This is a rote copy of the new suspQuad code in matgwinc that iterates over
the stages, and supports tapered fibers:

https://git.ligo.org/gwinc/matgwinc/commit/a59f01192a1ffc1644a45be536e31bc0eb78bd7c#9cb6e594e14c1d61dbb3c91549166a2d48db194f

This allows for setting per-stage temperature and wire and blade properties.

This version differs slightly in some generalizations about the number of
stages, although it doesn't quite yet support an arbitrary numer.

This is the scipy equivalent of the 'pchip' interpolation algorithm that
matgwinc is using.

Also minor clean up and clarify some of the variable names

Use function from R. Schilling CQG 14 1997.  Particularly needed for longer
arms.

better titles, calc Total difference, and show max fractional error in plots
and in log, and other log cleanup

also clean up docstrings.

Improves readbility.  use translation for MATLAB purposes.

Infrastructure is obviously where it really belongs.  This also allows us
to almost eliminate the need for the Constants struct on our side.

This is mostly used for MATLAB translation, but is useful for a couple of
other things as well (R_earth for instance).

These two functions were basically identical except for the material
used for the test mass stage, and a hacky reduction of the spring
constant for silicon.

suspQuad is modified to support different materials for the test mass
stage, and a hacky switch is added to adjust the spring constant.
suspBQuad is now just a wrapper around suspQuad specifying that the
test mass material is silicon instead of silica (there is definitely a
better way to do this, by e.g. specifying the fiber material in the
Stages struct list).

don't need to be defined inside of the suspQuad function

move this out of noise package, since it's not actually directly a noise
calculation.  no functional change.