
* Fix fudge bug fix in TWEAK_GAL code.  Also, apodize the ends of the input
  spectrum to minimize edge effects, or trim to points common to all test z's.

* Output synthetic ugriz colors for each object measured from the actual
  spectrum and from the synthesized spectrum (allowing u+z to be measured).

* Some z errors are negative!

* For PCA_SOLVE, trim to desired wavelength range before solving.

* For eigenspectra, save a 2nd HDU which is the remaining inverse-variance.

* Look at the e1/e0, e2,e0 values for the galaxies.

--------------DONE-------------------------------------------------------------
* Write PLATELIST routine to list reduced plates.

* For PCA_VSTND, get rid of bad endpoints made with too little data.

* Final correlation coeff should be computed with same wavelength-range
  for template, and masked + apodized the same as the object spectrum.

* Minimize reduced chi2 instead of chi2 ?

* Still need to compute errors in redshift, but how?

