check that pmAstromLinearizeTransforms does the right thing

[bills]

For distribution and postage stamps at the chip level we attempt
to do a linear fit to the bi-level astrometry.
We have had reports of errors. We need to insure that
these are just limitations of the fitting process and not bugs.

Also in streaksremove if the fit fails we leave the original wcs
in the chip and raw files. this might not be the best thing to do.

[eugene]

I believe these were fixed in r27461.

see also #1194