ncplane_mergedown() isn't working for small target planes

[dankamongmen]

Having added a background to the Tetris demo, I've made the playing area its own plane above the background. Attempting to ncplane_mergedown() settled pieces onto this plane, however, is running
into problems. I think this is due to continued use of lfdimx within paint() -- if paint() is our method of implementing ncplane_mergedown() (which it really does need to be), it oughtn't be using any parameters from notcurses* nc. Anywhere that it is is almost certainly a bug.

Don't call this "fixed" without a good unit test on it.

[dankamongmen]

I've added a unit test in dankamongmen/tetris, which is crashing, so huzzah!

[dankamongmen]

Oh weird this might be due to ncplane_polyfill_recurse()? According to valgrind anyway. That seems surprising:

==1390854== Command: ./notcurses-tester --tc=Fills --sc=MergeDownSmallPlane
==1390854== 
==1390854== Stack overflow in thread #1: can't grow stack to 0x1ffe801000
==1390854== Stack overflow in thread #1: can't grow stack to 0x1ffe801000
==1390854== Can't extend stack to 0x1ffe8010b8 during signal delivery for thread 1:
==1390854==   no stack segment
==1390854== 
==1390854== Process terminating with default action of signal 11 (SIGSEGV)
==1390854==  Access not within mapped region at address 0x1FFE8010B8
==1390854== Stack overflow in thread #1: can't grow stack to 0x1ffe801000
==1390854==    at 0x485F9F7: cell_simple_p (notcurses.h:1745)
==1390854==    by 0x48604CF: pool_release (internal.h:571)
==1390854==    by 0x486052A: cell_duplicate_far (internal.h:580)
==1390854==    by 0x4860955: cell_duplicate (render.c:95)
==1390854==    by 0x484F04C: ncplane_polyfill_recurse (fill.c:34)
==1390854==    by 0x484F07A: ncplane_polyfill_recurse (fill.c:38)
==1390854==    by 0x484F0AC: ncplane_polyfill_recurse (fill.c:42)
==1390854==    by 0x484F07A: ncplane_polyfill_recurse (fill.c:38)
==1390854==    by 0x484F0AC: ncplane_polyfill_recurse (fill.c:42)
==1390854==    by 0x484F07A: ncplane_polyfill_recurse (fill.c:38)
==1390854==    by 0x484F0AC: ncplane_polyfill_recurse (fill.c:42)
==1390854==    by 0x484F07A: ncplane_polyfill_recurse (fill.c:38)
==1390854==  If you believe this happened as a result of a stack
==1390854==  overflow in your program's main thread (unlikely but
==1390854==  possible), you can try to increase the size of the
==1390854==  main thread stack using the --main-stacksize= flag.
==1390854==  The main thread stack size used in this run was 8388608.
==1390854== Stack overflow in thread #1: can't grow stack to 0x1ffe801000
==1390854== 

[dankamongmen]

I ramped the polyfill unit tests up without trouble, so I think this might be a red herring.

[dankamongmen]

Ahhh, the polyfill() thing was that I was passing a null EGC, which yeah that won't work LOL. We ought guard against that. I've created #418 for this.

[dankamongmen]

Well shit, with #418 resolved, the unit test now passes :/. And yet things are not working in tetris...

[dankamongmen]

Resolved in dankamongmen/tetris. As predicted, we just needed to eliminate all uses of nc in paint(). I've done that, and the behavior in tetris is fixed.