fix(libSceAgcDriver): SPI_PS_INPUT_CNTL registers start at their clear-state value - #2128

What

A title that never writes SPI_PS_INPUT_CNTL_0 (context register 0x191) relies on its clear-state value, 0. The driver's initial context did not contain the SPI_PS_INPUT_CNTL registers. As a result, every draw of a pixel shader with interpolants (SPI_PS_IN_CONTROL.NUM_INTERP > 0) was rejected when [DecodePixelStageInfo](https://github.com/boykopovar/AnyPS5/blob/main/core/libs/prx/libSceAgcDriver/Graphics/src/ShaderInputState.cpp) read register 0x191.

Planet Coaster (PPSA01736) writes SPI_PS_INPUT_CNTL_1..31 (0x192..0x1b0, values 1..31) and leaves SPI_PS_INPUT_CNTL_0 at its reset value. Without this change it stops at its first UI and video draws:

[videoout] presentation failed: AGC graphics: missing register at DWORD 0x191

On older main the same draws were skipped instead ([gpu] skipped draw: AGC graphics: missing register at DWORD 0x191), and every presented frame was black.

All 32 SPI_PS_INPUT_CNTL registers (0x191..0x1b0) now start at 0 in [InitialContextRegisters](https://github.com/boykopovar/AnyPS5/blob/main/core/libs/prx/libSceAgcDriver/Execution/include/QueueState.hpp). The initial context and CLEAR_STATE already share that table.

Source for the value: the GFX10 clear state in amdgpu, clearstate_gfx10.h lines 427–458. In gfx10_SECT_CONTEXT_def_2, whose extent starts at 0xa0d8, SPI_PS_INPUT_CNTL_0 is entry 185. That is register 0xa191, the driver's 0x191. It and SPI_PS_INPUT_CNTL_1..31 are all 0x00000000.

Tested

Windows 11, MinGW-w64 15.2.0, RTX 3060.

• agc_driver_graphics_tests: a fresh queue state whose pixel stage has two interpolants and writes only SPI_PS_INPUT_CNTL_1 decodes CNTL_0 as 0 and CNTL_1 as written. ClearContext (CLEAR_STATE) resets all 32 to 0. Without the change, the test fails with missing register at DWORD 0x191.
• Full ctest suite: 494/494 passed. tools/check_conventions.py --base origin/main passes.
• PPSA01736: with the change, the PS5 notice screen and the Frontier intro video render. The title needs other fixes to get that far: fix(relinker): relink the modules that found needed modules need #1782, fix(libSceHttp): sceHttpWaitRequest reports no ready requests #1775, feat(libkernel): open /dev/random and /dev/urandom for reading #1777, fix(agc): defer shaders that cannot be prepared at registration to their first use #1836, feat(libkernel): implement mlock and munlock #1673 and some local commits. The log line above was taken from that build with only this change reverted.

Checklist

• Based on current main; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed in TechnicalDebt
• No notes, investigation .md files or images added to the repository
• New third-party code is a submodule built from source (none added)
• Generic behaviour, not specific to one title
• Depends on: none
• AI-assisted: yes