adjust CLOCK_CYCLES and related macros to handle fpu/iu concurrency

add CLOCK_CYCLES_FPU, which does exactly what CLOCK_CYCLES already did.

add CONCURRENCY_CYCLES, which sets fpu_cycles, which is the number of
available concurrent execution cycles that the integer unit can do
"free" work in while the fpu is executing.

adjust CLOCK_CYCLES so that if there are fpu_cycles, the cycle count is
subtracted from fpu_cycles instead of cycles, emulating the behavior of
these concurrent cycles being "free" as on real hardware.

src/cpu/808x.c

@@ -89,11 +89,32 @@ static int refresh = 0, cycdiff;
 		wait(val, 0);	\
 	}
 
-#define CLOCK_CYCLES(val)		\
+#define CLOCK_CYCLES_ALWAYS(val)		\
 	{			\
 		wait(val, 0);	\
 	}
 
+#define CLOCK_CYCLES_FPU(val)		\
+	{			\
+		wait(val, 0);	\
+	}
+
+
+#define CLOCK_CYCLES(val) \
+	{	\
+		if (fpu_cycles > 0) {	\
+			fpu_cycles -= (val);	\
+			if (fpu_cycles < 0) {	\
+				wait(val, 0);	\
+			}	\
+		} else {	\
+			wait(val, 0);	\
+		}	\
+	}
+
+#define CONCURRENCY_CYCLES(c) fpu_cycles = (c)
+
+
 typedef	int (*OpFn)(uint32_t fetchdat);