Add oam_spr_2x2 for drawing 16x16 sprites from sequential tiles

[Copilot]

Drawing a 2×2 (16×16) sprite currently requires 4 oam_spr calls with manual offset math. oam_spr_2x2 replaces this with a single call:

// Before: 4 calls with manual x+8, y+8 offsets and OAM slot math
oam_spr(x, y, 0xD8, 0, 0);
oam_spr((byte)(x + 8), y, 0xDA, 0, 4);
oam_spr(x, (byte)(y + 8), 0xD9, 0, 8);
oam_spr((byte)(x + 8), (byte)(y + 8), 0xDB, 0, 12);

// After
oam_spr_2x2(x, y, 0xD8, 0xD9, 0xDA, 0xDB, 0, 0);

Implementation

Implemented as a compile-time intrinsic — no new 6502 subroutine needed:

• NESLib.cs — Method stub: byte oam_spr_2x2(byte x, byte y, byte topLeft, byte bottomLeft, byte topRight, byte bottomRight, byte attr, byte sprid)
• IL2NESWriter.OamSprites.cs — EmitOamSpr2x2() handler that:
  • Backward-scans IL to classify 8 args (with Stloc skip logic for Roslyn Release optimizer reordering)
  • Builds a 17-byte metasprite array at compile time from tile/attr constants
  • Emits the existing oam_meta_spr calling convention (TEMP=x, TEMP2=y, PTR=array, A=sprid)
• IL2NESWriter.ILDispatch.cs — Dispatch routing
• Program6502.cs — Include oam_meta_spr subroutine when oam_spr_2x2 is used

Sample updates

• staticsprite — 4 oam_spr → 1 oam_spr_2x2 (all constant args)
• movingsprite — 4 oam_spr → 1 oam_spr_2x2 (local x/y, constant tiles)
• pong — Not updated (uses 1×3 paddles + single ball, not 2×2 patterns)

Tests

Two RoslynTests added: OamSpr2x2WithConstantArgs and OamSpr2x2WithLocalArgs.