Add support for simulating calibration exposures

[dkirkby]

From the updated simulate() docstring:

        Calibration exposures can be simulated by providing an array of
        ``calibration_surface_brightness`` values to use.  In this case,
        the source and fiberloss inputs are ignored, and no atmospheric
        emission or extinction are applied.

I added some examples to the bottom of the tutorial notebook.

[dkirkby]

Basic usage example:

sbunit = u.erg / (u.Angstrom * u.s * u.cm ** 2 * u.arcsec ** 2)
flat_lamp = 2e-13 * np.ones((desi.num_fibers, num_wlen)) * sbunit
desi.observation.exposure_time = 10 * u.s
desi.simulate(calibration_surface_brightness=flat_lamp, focal_positions=xy)

Calibration is implemented via:

• source_flux = surface_brightness * fiber_area
• fiberloss = 1
• sky emission = 0
• no sky extinction applied

Since the user provides a surface brightness for each fiber, arbitrarily complex illumination models are supported, but are not modeled within specsim. Vignetting should be implemented in specsim, and included here, but is not yet #56.

Fibers can be positioned either directly in (x,y) or via (ra, dec), as for non-calibration exposures.

[sbailey]

Looks great. Just what I need. Please merge when ready. Thanks.