Increase the frequency of CBRANCH

[tevador]

The main purpose of this change is to decrease the efficiency of VLIW-based custom hardware that uses static optimization techniques.

Definitions:

• branch span is the number of instructions that a CBRANCH instruction jumps backwards over
• trivial branch is a CBRANCH instruction that jumps back to itself, i.e. its span is 0

The effect of this change is following:

• the average number of branches per program increases from 16 to 25
• the average number of trivial branches per program increases from 2.1 to 4.0
• the average branch span decreases from 6.7 to 5.2
• the hashrate of Intel/AMD CPUs decreases by around 1% (sometimes there is no measurable difference)
• the theoretical IPC of a simulated VLIW design (using the Python script provided in the X41 audit) decreases by around 20%