Implementation is duplicated, but tolerate this for now, because there's
no clear idea how to de-dup it.

It's no longer intended to provide just "raw" socket interface, may include
some convenience methods for compatibility with CPython socket - but anyway
just minimal set required to deal with socket client and servers, not wider
network functionality.

Can't decide which is better for string type, char or byte pointer.
Changing to char removes a few casts.  Really need to do proper unicode.

This mirrors CPython behavior and makes possible to run scripts which
import other modules not from script's directory.

If it's not available, "~/.micropython/lib:/usr/lib/micropython" is used
as a fallback.

Both return int so far (single-precision float doesn't have enough
bits to represent int32 precisely).

Hopefully this works for other people as well :)

Proof of concept, controls "ffi" module as one which requires external
dependencies.

Foreign Function Interface module allows to load native dynamic libraries,
call functions and access variables in them. This makes possible to write
interface modules in pure Python.

This module provides thin wrapper around libffi. ctypes compatibility might
be possible to implement on top of this module (though ctypes allow to call
functions without prototypes, which is not supported by libffi (i.e.
implementation would be inefficient))).

-147 bytes on x86.

Also, don't try to force debug info format, use native.

TODO: Check lexer/parse/compile error path for leaks too.

Also, readline uses system malloc, so for symmetry, use the same for
non-readline case.

Can now have null bytes in strings.  Can define ROM qstrs per port using
qstrdefsport.h

CPython _socket actually have only those and doesn't provide stream
interface (higher-level CPython "socket" what adds this).

+516 bytes x86.

A file cannot be iterated concurrently, so we make io.FileIO its own
iterator.

Exceptions know source file, line and block name.

Also tidy up some debug printing functions and provide a global
flag to enable/disable them.

Byte code has a map from byte-code offset to source-code line number,
used to give better error messages.