This writes spectral images using the layout proposed in "An OpenEXR Layout
for Sepctral Images" by Fichet et al., https://jcgt.org/published/0010/03/01/.

Rename SampledWavelengths::SampleXYZ() -> SampleVisible()
And other minor renamings and formatting changes.

In the previous implementation, vignetting would affect the values of the
geometric quantities stored in the gbuffer, which is clearly not what
anyone wanted...

Fixes #149.

Rework interfaces for RealisticCamera exit pupil sampling and
providing the PixelSensor illuminant.

(This basically reverts eb353742.)

While this leads to an increase in color noise for scenes with dispersion,
it simplifies all of the integrators (and, for the book's pedagogical
purposes, lets us hold off on discussing this division until a better spot
in the book.)

Renamed RandomSampler -> IndependentSampler and thus, now it's
"independent" in scene description files rather than "random" to choose
that one.

Includes various other minor changes that don't affect functionality.

Multiple (mostly minor) changes:

- Time is removed from the ray weight computed by the RealisticCamera
  (it's handled by the PixelSensor).
- The "fnumber" and "c" parameters have been removed from the
  PixelSensor's parameters.
- The 683 scale has been removed from both the PixelSensor imaging ratio
  and from SpectrumToPhotometric() (so a no-op, mathematically.)
- Bugfix: Medium implementations keep a separate scalar scale for their
  emission, which is now normalized like regular light sources. This will
  cause image differences.

  diffs for scenes with volumetric emission.

Folded HaltonPixelIndexer back into HaltonSampler, among other small changes.

Remove variance from EXR metadata
Change GBufferFilm to compute per-RGB-component variances

Note: the Film's "scale" parameter has been removed, as it is redundant given "iso" now.

Add ToOutputRGB to FilmHandle and use it in SPPMIntegrator::Render to output in correct colour space

Previously, we'd add up contributions for all the vertices of a path and
then divide by wavengths' PDFs in Film::AddSample()/AddSplat(). The problem
with this was the case where secondary wavelengths are terminated in the
middle of path (e.g. for dispersion). What we actually should be doing (and
now are) is dividing by the current wavelength PDFs any time we add to the
radiance sum for a pth.  Thus, vertices before the secondary wavelengths
are terminated aren't weighted incorrectly, by the post-termination PDFs...

Remove sensor "exposuretime" parameter. Instead pass through time serived from camera shutter open/close times.

This is mostly various parts of Physlight moving around a bit in the code, but without changes to functionality.

Small bugfixes in Spectra::D() and the handling of Sensor in GBufferFilm.

Add Physlight.