Merge in the changesets that were missed:
- type recording
- better polymorphic IC performance
- build system improvements
- update to what was llvm trunk

Found some bugs in the name analyzer around listcomps; got the unittests back up and running so I can write some more

It all feel pretty hacked together though... definitely not
confident that I've gotten all of them.

I guess I had gotten those instructions from somewhere,
that said to just clone from trunk.
I ran into an issue installing gold through the instructions
and thought it might have been that, but I think it was the
bison issue I mention now in the instructions.
Probably better to leave it at a real release instead of tracking
trunk and having every run of the Pyston install giving a different
output.

Should support:
- multiple comprehensions
- multiple if conditions
- nested control flow expressions
- OSR'ing from the list comprehension

Though it tends to hit the OSR bug in the previous commit.

Some extra changes that could have been split out:
- use pointers-to-const instead of references-to-const for attribute-name passing,
  to make it harder to bind to a temporary name that will go away.
- add a 'cls_only' flag to getattr / getattrType to not have to special-case clsattrs
  (or simply get it wrong, in the case of getattrType)

Make the tester blow up if CPython exits with a negative exit code, which indicates some non-Python-level problem

Makes it qucker to run; fixes #20.

Also, set -Wno-extern-c-compat to silence libunwind-related warnings.
Fixes #27, hopefully for real this time.

This doesn't surface if you install libunwind globally since clang will
ignore errors from system libraries; this change should fix it for people
who are installing libunwind locally (as is the goal).
Fixes #27

In particular, lower all control-flow-involving expressions (such as boolops,
list comprehensions, if expressions) into actual control flow.  This requires
unfolding the AST into a flat structure, since the control-flow-expression can
be nested inside other expressions, including other control-flow-expressions.

The IRGenerator was doing this before, but was leading to duplication because
it's very similar to what the CFG generator already has to do.

Doing this earlier in the pipeline means that more analysis passes can work on
the lowered expressions, rather than having to be taught how they work.  In
particular, the name analysis does not need to have to special case the fact
that list comprehensions may set names -- but aren't guaranteed to if they don't
get executed or the iterator is empty.  I tried implementing list comprehensions
without doing this, and a bunch of the work was in reimplementing the logic that
the analyzers already implement.

As a side benefit, deopts become much easier since the AST is already unfolded;
this change gets rid of the messy fake-variable-setting way that the IRGenerator
was unfolding the AST to get this same effect.

fix text formatting 

Update README.md

List ccache as required in INSTALLING

Should fix #22.

Many args ir fix