agx: Account for mask when writing registers

To use fewer registers.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18687>

src/asahi/compiler/agx_register_allocate.c

@@ -41,8 +41,7 @@ agx_write_registers(agx_instr *I, unsigned d)
    case AGX_OPCODE_TEXTURE_LOAD:
    case AGX_OPCODE_TEXTURE_SAMPLE:
    case AGX_OPCODE_LD_TILE:
-      /* TODO: mask */
-      return 4 * size;
+      return util_bitcount(I->mask) * size;
 
    case AGX_OPCODE_LDCF:
       return 6;