Ugh.

Ignore new dropping from auto*

Ignore some new droppings in the po system.

Modernize the configuration a bit (for small values of modernize)
so avoid at least some obsolete macros.

Avoiding array formulas means that the delta engine can handle the
formulas leading to much faster convergence.

It used to be correct, but that was 10+ years ago.

Freeze the canvas groups while we create the initial views for sheet
objects.

See bug #692.

Performance problems were observed for copying a large number of
cell comments.  Part of that problem was measuring the pixel
width/height of large ranges of columns/rows.

The old design was linear leading to O(n^2) behaviour when used in
a loop.

This design caches the pixel position of each segment [block of
128 columns/rows].  With that, the amortized cost is constant.

It isn't clear that the current ColRowCollection design is right,
but this patch will lead a bit of life to the current design.

This moves colrow copying to sheet.c and we make that handle the creation
of the new colrow entry too.

We have too many near-copies of this code.  This consolidates some
of that.

Freeze the grid_items group also.  This group is used for cell comments
of which there can be a lot.

We may have lots of comments so determine styling once and hang it on
the pane.

If we see --pid=nnn, then don't add --args

We want round-away-from-zero (2.5 -> 3) not to-even (2.5 -> 2).
(This is for the case where either 2 or 3 will round back to the same
double -- see #638.)

Docs for this kind of thing either stink or are too well hidden.

We claim to be Excel compatible, but I haven't actually checked that.
There are lots of degrees of freedom and it doesn't matter.  Anyone
who complains will need to supply a working patch.

The g_set_prgname broke this.

Call g_set_prgname without path, except when help is requested.
See #677.

Use grep -E instead.