Add tests to check that we handle improperly encoded data

[jsdw]

#53 and #54 both avoid panics when the data isn't properly SCALE encoded. We should add a bunch of tests (perhaps using something like proptest) to check that we avoid panics with improperly encoded data.

[jsdw]

I added a couple of proptests to confirm that both of those issues are fixed in #56. We can extend this with more tests, though we do need to decide what type we think the bytes should be in the type info (eg String and BitVec for the above), so we need to be a bit more targeted than I was hoping here.

Maybe there is a more general fuzzer based approach that we can use to test combinations of all sorts of types in the metadata, or maybe we just add some tests for each of the other main SCALE types (sequences eg Vec, primitives, array, tuple, a couple of composites)