Less fragile subtree construction in parse transformation

[daniel-horpacsi]

#14 adds a parse transformation that changes message handling in programs under test. It works fine, but the subtrees are matched and constructed by using explicit data structures used in the representation, rather than via an API (see review comments here).

Unfortunately, this syntax tree representation doesn't seem to have a common API (unlike the format with erl_syntax), so in the worst case, we may introduce our own functions for the construct we use (see a similar example here).