So far, don't work for strings as expected.

We likely should make mp_obj_new_int() inline, and rely on its
encapsulated check rather than inline checks everywhere explicitly.

Also, parser for big small int values is still broken.

Currently it would report "operation not supported" which is confusing. Overall,
this is thinko leading to undefined behavior.

Still need to make built-ins by these names, and write tests.

Creating of classes (types) and instances is much more like CPython now.
You can use "type('name', (), {...})" to create classes.

Addresses issue #104.

It's not really about that, though; it's about me figuring out a sane
way forward for keyword-argument functions (and function
metadata). But it's useful as is, and shouldn't break any existing
code, so here you have it; I'm going to park it in my mind for a bit
while sorting out the rest of the dict branch.

Now much more inline with how CPython does types.

Qstr's are now split into a linked-list of qstr pools.  This has 2
benefits: the first pool can be in ROM (huge benefit, since we no longer
use RAM for the core qstrs), and subsequent pools use m_new for the next
pool instead of m_renew (thus avoiding a huge single table for all the
qstrs).

Still would be better to use a hash table, but this scheme takes us part
of the way (eventually convert the pools to hash tables).

Also fixed bug with import.

Also improved the way the module code is referenced (not magic number 1
anymore).

import works for simple cases.  Still work to do on finding the right
script, and setting globals/locals correctly when running an imported
function.