fix(agc): allow MSAA_ENABLE single-sample rect lists - #6563

Verified on PPSA03541 stacked with #6549 (local GDS+MSAA AgcDriver): past COPY_DATA GDS and past MSAA_ENABLE rect-list. Title reaches HDR present and issues rect-list draws (Rect-list draw of shader ... preparing it at draw). Still running past both stops; no abort on the previous MSAA path.

Other open pull requests touch the same code. Merging one will break the other, or both implement the same thing. If this one builds on them, list them in Depends on.

Some comparisons could not be completed because a pull request conflicts with the base branch. Review these shared files manually before concluding that there is no overlap.

Tested on Linux x86-64 (Release build, lavapipe / llvmpipe LLVM 20.1.2), with f9f67e3 merged into main 31cbb55:

• With the PR: all 8 tests built from Graphics.cpp pass.
• With main's sources: agc_driver_graphics and agc_driver_alpha_to_coverage_state fail with AGC graphics: unexpected failure: MSAA_ENABLE on a single-sample target for primitives other than triangles is unsupported (register 0x242 = 0x00000002).

AI-assisted: yes (Claude Code)

Rebase handoff — rebased onto current main cleanly (authorship unchanged).

Verified on Linux (CachyOS, kernel 7.2.9, GCC 16.2.1, Ninja, Release, BUILD_TESTING=ON, AMD RX 9070 XT / RADV):

• builds; tools/check_conventions.py --base origin/main passes
• full suite: 1338/1339 — the only failure is the known main regression (agc_driver_tessellation segfault, feat(shader): read hull inputs at a control-point index computed from data #4727), not this PR

To pick it up:

git fetch https://github.com/blawesom/AnyPS5 rebase/6563
git checkout <your branch>
git merge --ff-only FETCH_HEAD

Head is 04d023c07.

Tested on PPSA03541 (EA SPORTS UFC 5, content version 01.031.000): Windows 11, RTX 5090, NVIDIA 617.42, WinLibs GCC 15.2.0. After #6549, the title stopped here with "MSAA_ENABLE on a single-sample target for primitives other than triangles is unsupported (register 0x242 = 0x00000007)" about 20 s in. This PR cannot be applied to main as it is: main has been refactored, and nonTriangleRejection(queue, message, rectangles = false) in State.cpp now takes the rectangle case, which singleSampleDepthScanRejection already passes as true. I made the equivalent change on main 66702f1 (singleSampleMsaaRejection returns nonTriangleRejection(queue, "MSAA_ENABLE on a single-sample target for primitives other than triangles and rectangles", true)) and the title got past this stop. The branch here has conflicts that need that rebase. With #6549 and #6572 the title reaches the logos and main menu at about 34 fps; full results are in #7239.

AI-assisted: yes. Written with AI assistance and reviewed by me.

Main moved again (State.cpp was refactored around nonTriangleRejection), so this conflicted. I rebased onto current main and re-applied the change: singleSampleMsaaRejection now calls nonTriangleRejection(..., true), accepting VGT_PRIMITIVE_TYPE 7 and 17 like singleSampleDepthScanRejection. Points and lines still throw.

Added as our own commit on your branch:

• Graphics.cpp test: 2 and 3 are refused with "...other than triangles and rectangles"; 5, 6, 7, 17 and 21 are accepted.
• docs/dev/debt/libSceAgcDriver.md: the MSAA_ENABLE line now says filled triangles or rectangles are drawn and no longer lists rectangles among the throws.

Merging once the Build run is green. Thanks to @coltken95, @blawesom and @klosixgaming-arch for the runs.