Since 0be2ea50 axtls is automatically built
as part of the usual "make" build process.

Such names aren't stored as qstr in module dict, and there was a bug in
"import *" handling which assumed any name in a module dict is a qstr.

There was an assumption that all names in a module dict are qstr's.
However, they can be dynamically generated (by assigning to globals()),
and in case of a long name, it won't be a qstr. Handle this situation
properly, including taking care of not creating superfluous qstr's for
names starting with "_" (which aren't imported by "import *").

A board must be able to set the PLL3 values based on the HSE that it uses.

CPython does not have an implementation of select.poll() on some
operating systems (Windows, OSX depending on version) so skip the
test in those cases instead of failing it.

Taking the address of a local variable is mildly expensive, in code size
and stack usage.  So optimise scope_find_or_add_id() to not need to take a
pointer to the "added" variable, and instead take the kind to use for newly
added identifiers.

This ensures that implicit variables are only converted to implicit
closed-over variables (nonlocals) at the very end of the function scope.
If variables are closed-over when first used (read from, as was done prior
to this commit) then this can be incorrect because the variable may be
assigned to later on in the function which means they are just a plain
local, not closed over.

Fixes issue #4272.

The way it was written previously the variable x was not an implicit
nonlocal, it was just a normal local (but the compiler has a bug which
incorrectly makes it a nonlocal).

Building axtls gives a lot of warnings with -Wall enabled, and explicitly
disabling all of them cannot be done in a way compatible with gcc and
clang, and likely other compilers.  So just use -Wno-all to prevent all of
the extra warnings (in addition to the necessary -Wno-unused-parameter,
-Wno-uninitialized, -Wno-sign-compare and -Wno-old-style-definition).

Fixes issue #4182.

1. Use uctypes.bytearray_at().

Implementation of the "ffi" module predates that of "uctypes", so
initially some convenience functions to access memory were added
to ffi. Later, they landed in uctypes (which follows CPython's
ctype module).

So, replace undocumented experimental functions from ffi to
documented ones from uctypes.

2. Use more suitable type codes for arguments (e.g. "P" (const void*)
instead of "p" (void*).

3. Some better var naming.

4. Clarify some messages printed by the example.

To get rid of warning when building the docs saying there's a redirect from
http: to https:.

Examples are added to the beginning of the module docs, similarly to docs
for many other modules.

Improvements to grammar, style, and clarity. Some paragraphs are updated
with better suggestions. A warning added of the effect incorrect usage of
the module may have. Describe the fact that offset range used in one
defined structure is limited.

On almost all realistic platforms, native layout should be larger (or
equal) than packed layout.

sizeof() can work in two ways: a) calculate size of already instantiated
structure ("sizeof variable") - in this case we already no layout; b) size
of structure decsription ("sizeof type"). In the latter case, LAYOUT_NATIVE
was assumed, but there should possibility to calculate size for other
layouts too. So, with this patch, there're now 2 forms:

uctypes.sizeof(struct)
uctypes.sizeof(struct_desc, layout)

Configurable via MICROPY_MODULE_GETATTR, disabled by default.  Among other
things __getattr__ for modules can help to build lazy loading / code
unloading at runtime.

Configurable via MICROPY_PY_BUILTINS_STR_COUNT.  Default is enabled.
Disabled for bare-arm, minimal, unix-minimal and zephyr ports.  Disabling
it saves 408 bytes on x86.

Some Python linters don't like unconditional except clauses because they
catch SystemExit and KeyboardInterrupt, which usually is not the intended
behaviour.

This commit adds network.PPP(stream) which allows to create a TCP/IP
network interface over a stream object (eg a UART).

For example, on Android, pthread functions are part of libc, so LIBPTHREAD
should be empty.

Part of this test was trying to test some functionality of __getattribute__
but this method name was misspelt so it wasn't doing anything useful.
Fixing the typo in this name makes the test fail because MicroPython
doesn't support user defined __getattribute__ methods.  So this part of the
test is removed.  The remaining tests are modified slightly to make it
clearer what they are testing.

Any exception raised in a user __getattr__ should be propagated out.  A
test is added to verify these semantics.

Fixes issue #4116.

Fixes issue #4240.

1. Return correct error code for non-blocking vs timed out socket
(POSIX returns EAGAIN for both, we want ETIMEDOUT in case of timed
out socket). To achieve this, blocking/non-blocking flag is added
to the mp_obj_socket_t, to avoid issuing fcntl() syscall each time
EAGAIN occurs. (mp_obj_socket_t used to be 8 bytes, having some room
in a standard 16-byte alloc block.)

2. Handle socket.settimeout(0) properly - in Python, that means
non-blocking mode, but SO_RCVTIMEO/SO_SNDTIMEO of 0 is infinite
timeout.

3. Overall, make sure that socket.settimeout() call switches blocking
state as expected.

Prior to this commit the USB CDC used the USB start-of-frame (SOF) IRQ to
regularly check if buffered data needed to be sent out to the USB host.
This wasted resources (CPU, power) if no data needed to be sent.

This commit changes how the USB CDC transmits buffered data:
- When new data is first available to send the data is queued immediately
  on the USB IN endpoint, ready to be sent as soon as possible.
- Subsequent additions to the buffer (via usbd_cdc_try_tx()) will wait.
- When the low-level USB driver has finished sending out the data queued
  in the USB IN endpoint it calls usbd_cdc_tx_ready() which immediately
  queues any outstanding data, waiting for the next IN frame.

The benefits on this new approach are:
- SOF IRQ does not need to run continuously so device has a better chance
  to sleep for longer, and be more responsive to other IRQs.
- Because SOF IRQ is off, current consumption is reduced by a small amount,
  roughly 200uA when USB is connected (measured on PYBv1.0).
- CDC tx throughput (USB IN) on PYBv1.0 is about 2.3 faster (USB OUT is
  unchanged).
- When USB is connected, Python code that is executing is slightly faster
  because SOF IRQ no longer interrupts continuously.
- On F733 with USB HS, CDC tx throughput is about the same as prior to this
  commit.
- On F733 with USB HS, Python code is about 5% faster because of no SOF.

As part of this refactor, the serial port should no longer echo initial
characters when the serial port is first opened (this only used to happen
rarely on USB FS, but on USB HS is was more evident).

pyb.USB_VCP().isconnected() will now return False if the USB is
disconnected after having previously been connected.

See issue #4210.

So these constant objects can be loaded by dereferencing the REG_FUN_TABLE
pointer instead of loading immediate values.  This reduces the size of
generated native code (when such constants are used), and means that
pointers to these constants are no longer stored in the assembly code.

This shifts the work of loading the constant None object on to
load_reg_stack_imm(), making the handling of None more centralised.

This commit adds the helper function load_reg_stack_imm() which deals with
constant immediate values and converting them to Python objects if needed.