[feat] improve inst combine pass, optimize generated soufflé code

[ValKmjolnir]

inst combine pass will iterate multiple times to make sure necessary combinations are done.

for example:

( GF_4A_id(ssa_temp_2, a), ssa_temp_2 != 1, R_1A7__all__(ssa_temp_4), (ssa_temp_5 = ssa_temp_4, ssa_temp_4 = a), GF_4A_id(ssa_temp_6, ssa_temp_5), ssa_temp_6 != 1, 1 = 1 )

will be optimized to:

( GF_4A_id(ssa_temp_2, a), ssa_temp_2 != 1, R_1A7__all__(a), GF_4A_id(ssa_temp_6, a), ssa_temp_6 != 1 )

for ssa_temp_5 is replaced with ssa_temp_4 at first iteration, and ssa_temp_4 is replaced with a at next iteration.

[xiexie]

lgtm