kthread_create():  Rename kernel_thread() to kthread_create() for better naming consistency with task_create() and kthread_delete().

Initial ADC support for the STM32F33XX

* stm32_adc.h: add JEXTSEL definitions and hrtim trigger configuration

* stm32_adc.c: move STM32F33 ADC logic to a separate file

Approved-by: Gregory Nutt <gnutt@nuttx.org>

    All configurations that use NXIMAGE must select NX_MULTIUSER
    All configurations that use NXHELLO must select NX_MULTIUSER

6LoWPAN:  Fix a whole in the logic of the previous commit.  It turns out that g_uncomp_hdrlen has other usages so it cannot be modified as I was doing.  Instead, I needed to add a separate localt variable, protosize, to keep track of the two usages of g_uncomp_hdrlen.

stm32f7 BBSRAM: stm32_bbsram: avoid assert in stm32_bbsram_savepanic.  If panic happens before stm32_bbsram is initialized, stm32_bbsram_savepanic caused additional assert panic. Function has null pointer check, so drop DEBUGASSERT.

TCP Networking:  When CONFIG_NET_TCP_WRITE_BUFF=y there is a situation where a NULL pointer may be dereferenced.  In this configuration, the TCP connection's 'semi-permnanent' callback, s_sndcb was nullified in tcp_close_disconnect.  However, other logic in tcp_lost_connection() attempt to use that callback reference after it was nullifed.  Fixed in tcp_lost_connectino() by adding a NULL pointer change before the access.  This was reported by  Dmitriy Linikov in Bitbucket Issue 72.

STM32L4 small fixes to RTC

* STM32L4 RTC: init mode was never exited because nested locking in rtc_synchwait() disabled backup domain access

* STM32L4 RTC: use backup register magic value instead of INITS bit

    The INITS (bit 4) of RTC_ISR register cannot be used to reliably
    detect backup domain reset. This is because we can operate our
    device without ever initializing the year field in the RTC calendar
    if our application does not care about correct date being set.

    Hardware also clears the bit when RTC date is set back to year 2000:

    nsh> date -s "Jan 01 00:00:00 2001"
    rtc_dumptime: Setting time:
    rtc_dumptime:   tm: 2001-01-01 00:00:00
    rtc_dumpregs: New time setting:
    rtc_dumpregs:       TR: 00000000
    rtc_dumpregs:       DR: 00012101
    rtc_dumpregs:       CR: 00000000
    rtc_dumpregs:      ISR: 00000037
    ...
    nsh> date -s "Jan 01 00:00:00 2000"
    rtc_dumptime: Setting time:
    rtc_dumptime:   tm: 2000-01-01 00:00:00
    rtc_dumpregs: New time setting:
    rtc_dumpregs:       TR: 00000000
    rtc_dumpregs:       DR: 0000c101
    rtc_dumpregs:       CR: 00000000
    rtc_dumpregs:      ISR: 00000027      <--- Bit 4 went missing!
    ...

    This patch allows us to do:

      stm32l4_pmstop(true);

      /* Stop mode disables HSE/HSI/PLL and wake happens with default system
       * clock. So reconfigure clocks early on Stop mode return.
       */

      stm32l4_clockconfig();

    without stm32l4_clockconfig() doing spurious and harmful backup domain
    reset in rcc_resetbkp().

* STM32L4 RTC: put back the SSR race condition workaround

    ST has confirmed that the issue has not been fixed, and that it applies
    to STM32L4 too (was not in errata sheets due to documentation bug)
    See discussion:

    https://community.st.com/thread/43710-issue-with-rtc-maximum-time-resolution



* STM32F4, STM32L4, STM32F7 RTC: add more CONFIG_RTC_NALARMS > 1 to reduce code size

* STM32L4: rename stm32l4_rtcc.c to stm32l4_rtc.c to better match STM32F7

    Cosmetic changes to comments

* STM32, STM32L4, STM32F7 RTC: stray comment and typos in chip/stm32_rtcc.h

* STM32L4 RTC: change maximum alarm time from 24h to one month

Approved-by: Gregory Nutt <gnutt@nuttx.org>

fs/vfs:  Add new internal OS interface nx_read().  nx_read() is functionally equivalent to read() except that it does not modify the errno variable and it is not a cancellation point.  Changed all references to read() in the OS to nx_read().

net/socket: psock_send/psock_sendto: remove assert check for null psock and buf input pointers.  Removes check as 'psock == NULL' altogether because that checked for later in psock_send and psock_sendto. Change null check for 'buf' so that it is handled same as in recvfrom.c (return -EINVAL instead of
assert).