Improve speed

[fingolfin]

Running ward on src/c_oper1.c takes 8 seconds on my system (with LuaJIT, otherwise its more like 16 seconds).

It would be great if this could be sped up.

[fingolfin]

To quote @rbehrends:

the fallback case in the rewriter does not properly skip string literals

[rbehrends]

The underlying problem is that Ward actually goes to quite a bit of effort to eliminate spurious guards. This normally works well with manually written code, but the code generated by gac sometimes contains long functions with many variables, which drives up the cost of data flow analysis by a more than linear factor.