pkg/nimble: bump version to current master

gnrc_ipv6: don't recurse into receive for encapsulated IPv6

crypto: fix several documentation issues

Otherwise, an encapsulated IPv6 packet is handled twice. Once in the
central function, once in the specialized decapsulation.

`_decapsulate()` is called by callees of `_receive()` so the call to
the latter function within the first creates a recursion we don't want.
Using `gnrc_netapi` instead removes that and provides the added benefit
that other subscribers to IPv6 are also informed.

Update return values in documentation
Improve comments with separate @return statement for each rv
Remove incorrect return value for aes_init
Use @return instead of @returns
Fix comment lines over 80 chars

crypto: Fix code style

drivers/ethos: remove unnecessary casts.

ipv6_ext_rh: define type numbers at central place

Update Jiminy doc.txt

Added explanation for fuse settings.
Brown-out when 1.8V is set as system voltage.
Power increase in deep sleep when ext. xtal as system clock.

boards: Enable STM32F103C8 ROM hack via env var

emb6_sock_udp: copy receive remote correctly

test: trickle: fix condition for success

The `src` member of `sock_udp_t` for `emb6` is a pointer, not a value,
so it should not be referenced.

This fixes the output issue encountered during the
[2018.10 RC1 testing][RC1].

[RC1]: https://github.com/RIOT-OS/Release-Specs/issues/76#issuecomment-435924505

boards/msba2: Added call to periph_init

The current test implementation wrongly assumes that the diff between
two fired events (e1, e2) must always increase. That is not true, as
event e1 may reside on the upper part of [I/2, I) and e2 on the lower
part of [I, 2*I).

This commit fixes the test to look at the actual time that was randonmly
chosen from both intervals (t1, t2). Given that the intervals are
doubled, t1 must always be smaller than t2.

cpu/esp32: doxygen fix for periph/gpio

cpu/esp8266: doxygen fix for periph/gpio

The STM32F103C8 secretly comes with 128KiB flash instead of 64KiB. Still, only
64KiB of it are tested and guaranteed to work. However, most of the times the
whole 128KiB flash works just fine. In the BluePill documentation this fact is
already documented and by using

    $ make BOARD=bluepill CPU_MODEL=stm32f103cb

the whole 128 KiB can be used by RIOT. When using this hack routinely, it easier
to use environment variables instead. But allowing to overwrite CPU_MODEL via
environment variables seems to be a bad thing, as it is easy to forget to clear
that environment variable when changing the BOARD variable.

This commit introduces the new STM32F103C8_FLASH_HACK variable, which unlocks
the 128KiB FLASH when set to "1". The BluePill documentation has been updated
accordingly.

sys/auto_init: Fixed initialization of sht1x [backport 2018.10]

Removes architecture specific includes from documentation of module  Peripheral Driver Interface / GPIO.

Removes architecture specific includes from documentation of module  Peripheral Driver Interface / GPIO.

net/nanocoap: make data pointer c++ compliant

cpu/atmega*: factorise common code into atmega_common

The unification of a bigger stack for the atmega platforms
makes some boards to not have enough memory to provide
the big stack plus the application code.

It is possible though, to override the stack size to a
smaller amount if running the test is necessary.

This variable helps to inform at boot time that some
information about the booting process, e.g. reset cause.

Empty array uint8_t data[] is not allowed in ISO-C++. Replacement: function coap_hdr_data_ptr, which handles
the pointer arithmetic to point where hdr.data pointed