m_malloc already checks for a failed allocation so there's no need to check
for it in m_malloc0.

And it's no longer unconditionally included by nlr.h, only if NLR_SETJMP
is defined.

Each NLR implementation (Thumb, x86, x64, xtensa, setjmp) duplicates a lot
of the NLR code, specifically that dealing with pushing and popping the NLR
pointer to maintain the linked-list of NLR buffers.  This patch factors all
of that code out of the specific implementations into generic functions in
nlr.c.  This eliminates duplicated code.

The factoring also allows to make the machine-specific NLR code pure
assembler code, thus allowing nlrthumb.c to use naked function attributes
in the correct way (naked functions can only have basic inline assembler
code in them).

There is a small overhead introduced (typically 1 machine instruction)
because now the generic nlr_jump() must call nlr_jump_tail() rather than
them being one combined function.

Where possible it's important to test all code in the code base.

In this case the result should always be 0, even if 2nd arg is 0.

set_equal is called only from set_binary_op, and this guarantees that the
second arg to set_equal is always a set or frozenset.  So there is no need
to do a further check.

Breaks up HW SPI transfers into maximum chunks of 32736 bits (4092 bytes),
because this is the maximum that the underlying ESP IDF will accept.

Previously, testing of stackless build happened (manually) in
travis-stackless branch. However, stackless offers important
featureset, so it's worth to test it as a part of the main
CI. Strict stackless is used because it's the "real" stackless
build, which avoids using C stack as much as possible (non-strict
just prefers heap over C stack, but may end up using the latter).

This build sets CFLAGS_EXTRA itself, but preserve user's value as passed
on make command line/etc.

CPython 3.6 contains some backward incompatible changes, and further
version(s) are expected to have more. As we anyway implemente 3.4 with
some features of 3.5, refer to 3.5 docs to avoid confusion.

Examples of 3.6 backward incompatibilities:

https://docs.python.org/3.6/library/json.html#json.dump
https://docs.python.org/3.6/library/json.html#json.load

> Changed in version 3.6: All optional parameters are now keyword-only.

https://docs.python.org/3.6/library/functions.html#type

> Changed in version 3.6: Subclasses of type which don’t override
> type.__new__ may no longer use the one-argument form to get the
> type of an object.

https://docs.python.org/3.6/library/collections.html#collections.namedtuple

> Changed in version 3.6: The verbose and rename parameters became
> keyword-only arguments.

This implements .pend_throw(exc) method, which sets up an exception to be
triggered on the next call to generator's .__next__() or .send() method.
This is unlike .throw(), which immediately starts to execute the generator
to process the exception. This effectively adds Future-like capabilities
to generator protocol (exception will be raised in the future).

The need for such a method arised to implement uasyncio wait_for() function
efficiently (its behavior is clearly "Future" like, and normally would
require to introduce an expensive Future wrapper around all native
couroutines, like upstream asyncio does).

py/objgenerator: pend_throw: Return previous pended value.

This effectively allows to store an additional value (not necessary an
exception) in a coroutine while it's not being executed. uasyncio has
exactly this usecase: to mark a coro waiting in I/O queue (and thus
not executed in the normal scheduling queue), for the purpose of
implementing wait_for() function (cancellation of such waiting coro
by a timeout).

tinytest is written with the idea that tests won't write to stdout, so it
prints test name witjout newline, then executes test, then writes status.
But MicroPython tests write to stdout, so the test output becomes a mess.
So, instead print it like:

    # starting basics/andor.py
    ... test output ...
    basics/andor.py: OK

If TEST is defined, file it refers to will be used as the testsuite
source (should be generated with tools/tinytest-codegen.py).

"make-bin-testsuite" script is introduce to build such a binary.

The whole idea of --list-tests is that we prepare a list of tests to run
later, and currently don't have a connection to target board. Similarly
for --write-exp - only "python3" binary would be required for this operation,
not "micropython".

Way to reproduce a UnicodeEncodeError without this patch:

    LC_ALL=C tinytest-codegen.py ...

Some compilers can treat enum types as signed, in which case 3 bits is not
enough to encode all mp_raw_code_kind_t values.  So change the type to
mp_uint_t.

Everything happens automagically with overridden "sdist" from
sdist_upip.py.

If we test for unix target, do that explicitly. pyb var will be None
for commands like --list-tests too.

Instead of passing thru more and more options from tinytest-codegen to
run-tests --list-tests, pipe output of run-tests --list-tests into
tinytest-codegen.

The idea that --list-tests would be enough to produce list of tests for
tinytest-codegen didn't work, because normal run-tests processing heavily
relies on dynamic target capabilities discovery, and test filtering happens
as the result of that.

So, approach the issue from different end - allow to specify arbitrary
filtering criteria as run-tests arguments. This way, specific filters
will be still hardcoded, but at least on a particular target's side,
instead of constant patching tinytest-codegen and/or run-tests.

"skip <test>" message could leak before.

It relies on MICROPY_CPYTHON_COMPAT being defined.