[daip] FITS header

[Eric Greisen]

I am revising the subroutine that interprets headers with CDi_j and
PCi_j.  Previously it simply took CD1_1/cod(rot) as axis 1 increment
and CD2_2/cos(rot) as axis two increment.  Now it averages in the
CD2_1/sin(rot) and -CD1_2/sin(rot) estimates and prints a warning if
they differ even by a little.

Note that the new code brings the increments closer together but
still leaves them different by about 2%.  It prints a warning in the
case you provided.

This is still not what ds9 appears to be doing.  That program seems to
have deduced that the two increments have the same value (except for
sign).  I managed to test another pixel well separated from the
previous one and that is consistent with what aips gives with a
forcibly revised header.

Eric Greisen