Fixes issue #840

Declare do_str() function before the implementation

This ensures that GCC does not discard the do_str implementation in
some cases eg when compiling tests with debug enabled:
  make RUN_TESTS=1 DEBUG=1

Allows to create socket objects that support TCP and UDP in server and
client mode.  Interface is very close to standard Python socket class,
except bind and accept do not work the same (due to hardware not
supporting them in the usual way).

Not compiled by default.  To compile this module, use:
make MICROPY_PY_WIZNET5K=1

Update teensy README.md file

Thanks to Artur Wroblewski for some suggested changes.
I also added the TIPs section at the end while I was updating.

Part of code cleanup, working towards resolving issue #50.

Part of code cleanup, working towards resolving issue #50.

Part of code cleanup, working towards resolving issue #50.

Part of code cleanup, to resolve issue #50.

Part of code cleanup, towards resolving issue #50.

Addressing issue #50, still some way to go yet.

Top-level lib directory is for standard C libraries that we want to
provide our own versions of (for efficiency and stand-alone reasons).
It currently has libm in it for math functions.

Also add atanf and atan2f, which addresses issue #837.

Found this bug by running unix/ tests with DEBUG=1 enabled when
compiling.

Addresses issue #838.

Optimises:
    if () -> if False
    if (x,...) -> if True
    if (a and b) -> if a and b

This way, the native glue code is only compiled if native code is
enabled (which makes complete sense; thanks to Paul Sokolovsky for
the idea).

Should fix issue #834.

Basic native ARM emitter

The heap allocation is now exactly as it was before the "faster gc
alloc" patch, but it's still nearly as fast.  It is fixed by being
careful to always update the "last free block" pointer whenever the heap
changes (eg free or realloc).

Tested on all tests by enabling EXTENSIVE_HEAP_PROFILING in py/gc.c:
old and new allocator have exactly the same behaviour, just the new one
is much faster.

Recent speed up of GC allocation made the GC have a fragmented heap.
This patch restores "original fragmentation behaviour" whilst still
retaining relatively fast allocation.  This patch works because there is
always going to be a single block allocated now and then, which advances
the gc_last_free_atb_index pointer often enough so that the whole heap
doesn't need scanning.

Should address issue #836.