Includes a fix for H7 DAC DMA requests.

This turns a hard crash in a recursive generator into a 'maximum recursion
depth exceeded' exception.

The 2nd and 3rd args of the ternary operator are treated like they are in
the same expression and must have similar types.  void is not compatible
with int so that's why the compiler is complaining.

This patch moves the implementation of stream closure from a dedicated
method to the ioctl of the stream protocol, for each type that implements
closing.  The benefits of this are:

1. Rounds out the stream ioctl function, which already includes flush,
   seek and poll (among other things).

2. Makes calling mp_stream_close() on an object slightly more efficient
   because it now no longer needs to lookup the close method and call it,
   rather it just delegates straight to the ioctl function (if it exists).

3. Reduces code size and allows future types that implement the stream
   protocol to be smaller because they don't need a dedicated close method.

Code size reduction is around 200 bytes smaller for x86 archs and around
30 bytes smaller for the bare-metal archs.

The LHS passed to mp_obj_int_binary_op() will always be an integer, either
a small int or a big int, so the test for this type doesn't need to include
an "other, unsupported type" case.

The string "Q+?" is special in that it hashes to zero with the djb2
algorithm (among other strings), and a zero hash should be incremented to a
hash of 1.

Without the compiler enabled the mp_optimise_value is unused, and the
micropython.opt_level() function is not useful, so exclude these from the
build to save RAM and code size.

So that these tests test the runtime behaviour, not the compiler (which may
be executed offline).

The case of a return statement in the try suite of a try-except statement
was previously only tested by builtin_compile.py, and only then in the part
of this test which checked for the existence of the compile builtin.  So
this patch adds an explicit unit test for this case.

When pystack is enabled mp_obj_fun_bc_prepare_codestate() will always
return a valid pointer, and if there is no more pystack available then it
will raise an exception (a RuntimeError).  So having pystack enabled with
stackless enabled automatically gives strict stackless mode.  There is
therefore no need to have code for strict stackless mode when pystack is
enabled, and this patch optimises the VM for such a case.

Neither the ip nor sp variables are used again after the execution of the
RAISE_VARARGS opcode, so they don't need to be updated.

Some code in mp_iternext() was only tested by the native emitter, so the
tests added here test this function using just the bytecode emitter.

The VM expects that, if mp_resume() returns MP_VM_RETURN_EXCEPTION, then
the returned value is an exception instance (eg to add a traceback to it).
It's possible that a value passed to a generator's throw() is not an
exception so must be explicitly checked for if the thrown value is not
intercepted by the generator.

Thanks to @jepler for finding the bug.

Prior to this patch the code would crash if a key in a ** dict was anything
other than a str or qstr.  This is because mp_setup_code_state() assumes
that keys in kwargs are qstrs (for efficiency).

Thanks to @jepler for finding the bug.

Instead of using a dedicated variable in RAM it's simpler to use the
relevant bits in the DWT register.

If the board is configured to use a bootloader then that bootloader will
pass through the reset_mode.

The main() function has a predefined type in C which is not so useful for
embedded contexts.  This patch renames main() to stm32_main() so we can
define our own type signature for this function.  The type signature is
defined to have a single argument which is the "reset_mode" and is passed
through as r0 from Reset_Handler.  This allows, for example, a bootloader
to pass through information into the main application.

The Reset_Handler needs to copy the data section and zero the BSS, and
these operations should be as optimised as possible to reduce start up
time.  The versions provided in this patch are about 2x faster (on a Cortex
M4) than the previous implementations.

Rather than pin objects themselves.  The actual object is now pin_X_obj and
defines are provided so that pin_X is &pin_X_obj.  This makes it so that
code that uses pin objects doesn't need to know if they are literals or
objects (that need pointers taken) or something else.  They are just
entities that can be passed to the map_hal_pin_xxx functions.  This mirrors
how the core handles constant objects (eg mp_const_none which is
&mp_const_none_obj) and allows for the possibility of different
implementations of the pin layer.

For example, prior to this patch there was the following:

    extern const pin_obj_t pin_A0;
    #define pyb_pin_X1 pin_A0
    ...
    mp_hal_pin_high(&pin_A0);

and now there is:

    extern const pin_obj_t pin_A0_obj;
    #define pin_A0 (&pin_A0_obj)
    #define pyb_pin_X1 pin_A0
    ...
    mp_hal_pin_high(pin_A0);

This patch should have minimal effect on board configuration files.  The
only change that may be needed is if a board has .c files that configure
pins.

The H7 SD peripheral has direct connection to MDMA instead.

The relevant common.ld file should now be included explicitly by a
particular board.