compiler: improve boundary analysis and optimizations

[mertcandav]

This PR implements better boundary analysis to optimize memory accesses.

In current implementation, boundary analysis was only performed for iterations, to do this the semantic analyzer associated each iteration variable with the iteration expression and had some strict conditions: for example optimization was mostly applied when the index value was immutable, but it could not optimize the cost of boundary checking where it was mutable and could be used safely before being mutated.

In addition, compiler optimizes array indexing with constant expressions. However, that's a simple optimization. Just check whether expression is constant and fits size of array.

New Boundary Analysis Implementation
The new boundary analysis tracks constant values ​​and expressions that can be predicted at runtime. If the expressions are safe until they are mutated or being risky, it optimizes boundary access checking and provides faster access to memory at runtime without compromising safety.

For example:

len(x) == 3 && x[0] == 'f' && x[1] == 'o' && x[2] == 'o'

In the code example above, the compiler knows at compile time that the length of x is 3, so it skips the runtime boundary checking analysis of subsequent accesses.