polynomial fitting routines do not consistently implement coeff masks

[eugene]

1D and 2D ordinary polynomial fitting was not respecting the coefficient masks
(3D and 4D were). I have fixed this and included the change in the rel10_ifa
branch.

Also, the coefficient errors are not being correctly calculated. The
coefficient error is the square root of the inverse of the A matrix used to
calculate the coefficients (in the equation Ax = B). The GaussJordan method of
solving Ax=B leaves the inverse of A behind in A, but the LUD method does not.
either the GaussJordan method should be used or the LUD method needs to solve
for inverse(A) as part of its analysis. In the version checked into rel10_ifa,
I am using the GaussJordan version. (Perhaps that should be #ifdef'ed, not if
(1) {})

[David.Robbins@…]

done.

[Paul Price]

Bug has been resolved.... closing.

[Paul Price]

Bugs have been fixed... closing.