Fixes for issues #477 and #478

[mihaibudiu]

This fixes issue #477 and #478 but discovers several other bugs in the process.
Issues have been filed for these bugs. Unfortunately this also converts several tests to XFAIL.
An issue has been filed against the spec: p4lang/p4-spec#161; the current checks are quite strict, and so several other test programs have to be modified to pass, including some gtestp4c tests.

In the end the checks are part of the typechecking pass - it was lower overhead than creating a separate pass.

I have also converted more comments to Doxygen.

[ChrisDodd]

This feels a little bit like it should be a separate pass to me, rather than lumped in with TypeInference, but its ok as is.

[mihaibudiu]

That's how I wrote it originally, but then I realized it needs fresh types, so it will need to run typechecking again... We could split the type-checking pass into a few smaller passes, notably @cc10512 said he would like to move the compile-time computation to a separate pass too.