Improve blame error part 3

[yannham]

Depends on #141. Follow-up of #140 and #141. Filter and show the callstack when an higher-order contract fails.

what it does

• process the call stack to group elements into atomic calls and filter out calls to standard contracts
• show the callstack when a higher-order blame error occurs

limitations
An unfortunate limitation is that codespan does not preserve the original order of labels. That is, if we just show the successive calls as labels of the original error, they will be re-ordered and possibly fused depending on their respective position in the source, losing the temporal information, as in the following example:

Here, the calls happen in the order (1)-(2)-(3), but are displayed in the order (2)-(3)-(1).

Possible fixes are:

1. Use a distinct diagnostic to display each call of the callstack.
2. Switch to a different error reporting library which gives some control on the order of error message.
3. Have a separate rendering engine for the call stack, but this means more or less reimplementing half of codespan.
4. Hack into codespan to make it possible to preserve the ordering of labels.

I went for solution 1. for the time being, which requires minimal changes while the others have heavy implications. It gives reasonable results, although there are these annoying "note:" headers, as showed below:

[dpulls]

🎉 All dependencies have been resolved !

[aspiwack]

The “note” label is not that bad, if not completely obvious.

I don't know whether the stack filtering is the right way to produce a useful call stack (it seems to give decent results at least). But the documentation is very good, so it'll be clear how to alter it in the future, should the need arise.