notcurses_at_yx() doesn't return EGC in cell->gcluster, only in return value

[dankamongmen]

notcurses_at_yx() faithfully reproduces the reflected EGC in its return value, but rather than duplicating it in the value-result cell, sets that cell's gcluster to 0. this seems rather niggardly. can we not afford the short-term duplication into the standard plane's egcpool? Maybe not, if this is used in some tight loop somewhere in the core, but I very much doubt that to be the case outside of notcurses_refresh().

If we decide we can't suffer this short-lived copy, that's cool, hack on, but in that case we ought instead of V-R on the cell, we ought V-R on an attrword and a channels. The annoying thing about the current situation is that if you go working with the cell, i.e. passing it to is_simple() etc., it's broken due to that 0'd gcluster. So that's gotta go.

[dankamongmen]

notcurses_at_yx() has the following comment:

        egc = pool_egc_copy(&nc->pool, srccell);
        c->gcluster = 0; // otherwise cell_release() will blow up

and this makes sense--as opposed to ncplane_at_yx() (which does leave the EGC in ->gcluster), we don't have a good egcpool to stick this in. OK, fine.

In that case, eliminate notcurses_at_yx(), since it is surprising, and instead take the attrword and channels as value-results.