ABI interface fixes

[iamdefinitelyahuman]

What I did

Fix issues with bytes, string and fixed168x10 when importing an ABI as an interface. Fixes #1832

How I did it

1. When determining if an interface has been implemented, ignore the length on bytes and string.
2. Substitute fixed168x10 for decimal when converting ABI to AST.
3. For bytes and string arrays, use a length of 1024*1024 in inputs, 1 in outputs. The compiler does not implicitly truncate values so this does not create an issue when a limit is exceeded.

How to verify it

Run the tests. I expanded the cases around ABIs as interfaces to check each type as an input and output.

Cute Animal Picture

[michwill]

Excellent, does it also fix the same issue with structs which contain uint256 arrays?

[michwill]

E.g. this:

Z256: constant(uint256) = 0 

struct A:
    many: uint256[3]
    one: uint256

@private
@constant
def foo(_many: uint256[3], _one: uint256) -> A:
    a: A = A({many: [Z256, Z256, Z256], one: 0})
    a.many = _many
    a.one = _one
    return a

@public
def bar():
    m: uint256[3] = [Z256, Z256, Z256]
    m[0] = 42
    m[1] = 2020
    m[2] = 21000000
    out: A = foo(m, 33)
    assert out.many[0] == m[0]
    assert out.many[1] == m[1]
    assert out.many[2] == m[2]
    assert out.one == 33

[iamdefinitelyahuman]

@michwill good call, I'll have a look at that as well

[michwill]

As far as I can tell, this PR doesn't fix it for structs, and probably the solution would be suitable for this PR, too

[fubuloubu]

Overall, not really a big fan of this approach, but if there's no better way to solve the problem, the least we can do is add more notes to the code about it.

[iamdefinitelyahuman]

note to self, need to open another issue re: structs in json interfaces