Christopher Wipf authored Mar 10, 2020 and Christopher Wipf committed Mar 10, 2020

Responding to some MR !61 comments:
- docstrings were improved
- most bottom-stage special cases were rolled into new helper functions (`wireGeometry`, `wireTELoss`, `continuumWireK{h,v}`)
- suspensionthermal.py uses `zeros_like` where appropriate
- comment arithmetic was corrected in the Voyager parameters
- `TempLower` and `TempUpper` are returned correctly by `getJointParams`
- `material` kw argument to suspQuad was removed (again)
- `fullylossy`/`singlylossy` outputs was removed (again)
- dilution factor was checked (I think it's correct: see ref GG sec. 2.5)
- vertical spring constant from ref GG was checked (I think it's correct: GG assumed a 2-wire suspension)

Support was added for top stage temperature gradients.  sus.Temp, if defined, is assumed to be the temperature of the topmost joint.

Thermal noise is now computed by summing over the noise of each individual joint (no more "temperature regions").

Support for the stage.FiberRadius parameter has been removed.  None of the canonical IFOs use this.  If anyone ever used it, they can switch to stage.WireRadius instead.

The "questionable approximation" removed in the previous commit was reverted.  It should now agree with what is in master, so that tests pass whenever the suspension temperature is uniform.  This was done to keep the MR focused on restructuring and temperature-gradient related issues.  This approximation (and any other separable issues) can be addressed in another MR.

This commit does not yet deal with Kevin's comment on the splitting of the bottom stage noise.  I would like to find a way to address that while also allowing for a temperature gradient in the bottom stage.

I suggest opening new bugs for Kevin's comments on the vertical TE geometric factors and tapered fiber approximations.