AST: Python Abstract Synatx Tree
(S)CFG: (Structured) Control Flow Graph

Items to note for reviewers:

* Most of changes are in the `ast_transforms.py` module which contains
  the meat of the implementation.
* Temporary data structures are created for a pure AST based CFG
  (`ASTCFG`) with blocks that are writable. This made it much easier to
  implement the builder pattern. They can easily be converted to a SCFG
  using the `ASTCFG.to_SCFG()` function.
* The `AST2SCFGTranformer` is stateful and is populated with
  housekeeping data structures post transform.
* The tests are all in `test_ast_transforms.py`. They are of the usual
  format: Python function as input and serialized graph to compare to.
* The tests also check the values of the unreachable and empty nodes
  that have been pruned from the CFG.
* The for-loop decomposition is a bit involved, hence there is an
  extensive docstring to describe what it does.
* Some minor changes had to be made to the `scfg.py` module to support
  displaying "broken" SCFGs during development. This makes it easier to
  visually debug the graph upon making programming mistakes.
* The modules `basic_block.py` and `rendering.py` were modified to
  support construction and visualization of the `PythonASTBlock`, which
  is a new block type for the SCFG that contains a block derived from
  Python AST.
* The central entrypoint is: `AST2CFG(function) -> SCFG `. This will
  turn a Python function into an SCFG.
* A stub has been added for the invserse, SCFG -> AST transformer.

Example:

```python
from numba_rvsdg import AST2SCFG
def fun():
    ...
scfg = AST2SCFG(fun)
scfg.render()
scfg.restructure()
scfg.render()
```

As title

As title

As title

As title

This reduces the graph even more, since blocks with a single break or continue become empty blocks and are pruned

As title

The jump_targets for the else-branch of the for-loop must be set after
recursion, not before. Otherwise nested flow-control constructs in the
else-branch will fail to codegen correctly, leading to a broken CFG with
blocks that point to non-existing blocks. The test exposes this and
demonstrates the fix is valid.

This test exposes a bug in a missing seal for using continue in the
nested for-else clause. Specifically, one needs to have an else-clause
with a series of linear statements and a continue to hit this. If you
try to break, the break index will be the same as the exit index for the
seal_inside_of_loop as the default_index is also the exit_index. So you
can only hit this with exit_index. In that case the block with the
continue does not point to the loop header but to the incorrect exit
instead.

The test has been constructed such that a potential reconstruction would
also fail execution with an incorrect return value.

As title

As title

As title

As title

As title

As title

As title

As title

As title

As title

As title

As title

As title