Fix uses of addGuard(python_object)

Since we are putting more things in the IC gc references,
call GC again after clearing them.

Hopefully this doesn't end up keeping things alive for too much
extra time.

and add the test that exposed it.

addGuard(obj) is dangerous because obj could be deallocated+reallocated
in the same spot, which is what ended up happening here.

VRegInfo: reduce memory usage

bjit: fix memory corruption

we used to destroy the ICInfo before all references to it got deleted.

I run into this when using the glibc alloc instead of jemalloc

SourceInfo: remove the AST stmt copy, remove special handling of lambdas

had to change AST_Expression to store a stmt instead of a expr but this has the advantage that memory managment is easier

use PyObject_Malloc in pyston::DenseMap, smaller ICInvalidator, reduce malloc calls

llvm::SmallPtrSet<ICSlotInfo*, 2> and std::unordered_set<ICSlotInfo*> have the same size
but the SmallPtrSet has the advantage that it can already store 2 elements in there and does not allocate 10 elements in the constructor

it is slighlty faster for small allocations and has the advantage that it is easier to track how much memory our python objects are using

runtime ICs: use mmap instead of malloc for executable code

don't handle functions containing lambdas with yields as generators

make inner links to dir format

I'm suprised we did not run into this earlier...

use more often std::unique_ptr

This stuff got mostly already manually freed but I think this change makes it a little easier to see who owns what.

we should not execute memory which got allocated using malloc
(crashes a lot of memory inspection tools and also the windows subsystem for linux)
I replaced it with a trivial mmap allocator which never unmaps the memory
(but reuses the memory if a runtime ICs gets deleted)
I think this is fine for now:
- there are <100 runtime ICs used per test
- because all runtime ICs have the same size there is no problem with fragmentation

ICSlotInfo: remove old invalidator entries

delete the llvm module after code generation

this saves a lot of memory

Switch to CPython list sort(Not its list implementation)

If BoxedList::allocated is -1, it means the items inside were changed.
Some CPython list functions need this to check exceptions.

CPython listsort need "allocated" to check and throw exception:
https://github.com/Daetalus/pyston/blob/d84105ffc8a4855bb6e00d9c22ba2baa8bddf969/from_cpython/Objects/listobject.c#L2091
https://github.com/Daetalus/pyston/blob/d84105ffc8a4855bb6e00d9c22ba2baa8bddf969/from_cpython/Objects/listobject.c#L2180

Comment out some part of listobject.c, use the CPython list sort and
apply some changes to existed Pyston code.

ScopeNameUsage merge dicts into a single big one

and change it to use a unordered_map because it uses less memory in this case and is faster
(I assume because it does not have align the key value tuples)
saves about 10% of peak memory on django

Update to fewer-pyston-changes virtualenv