DrawImage: stop re-proving what a repeat already proved, and keep rowOpaque out of line - #13

…Opaque out of line

Two defects, both found by measuring the widgets rather than the
primitive, and both invisible from reading the code.

v0.7.0 regressed the unclipped cases it was not meant to touch: a full
1000x700 blit went from 265,039 ns/op at v0.5.0 to 1,152,535. The clipped
rewrite made the loop bigger, and rowOpaque -- a tight strided scan whose
cost lives entirely in register allocation -- was inlined into it. Kept
out of line the SAME code measures 238,252. The directive carries that
pair of numbers, because a reader deleting it deserves to know what it
costs. Profiling pointed at rowOpaque while attributing memmove's work to
it; only crossing the two implementations against each other, then the
two inlining choices, settled which of the two was to blame.

Separately, an enlarged blit scanned every source row TWICE: once when
the row was built, once for each destination row repeating it -- 700
scans of 1.4 MB per blit to prove 350 rows opaque. Settling the repeat
before touching the source removes that half outright. prevSY only ever
holds a row that was written whole, so a match already proves opacity.

  1:1        279726 ns/op  (per-pixel 1867584)   6.7x
  enlarged   351219 ns/op  (per-pixel 2066103)   5.9x
  clipped    314708 ns/op  (per-pixel 2644680)   8.4x

Every shape is now faster than it was at any earlier version: v0.6.0 blit
scaled in 434,038 and v0.7.0 blit clipped in 911,278.

100% statement coverage; the 16-case comparison against the loop
DrawImage replaced is unchanged and still passes.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>