May happen with a faulty configuration (eg no allowed curve but trying to use
ECDHE key exchange), but not trigger able remotely.

(Found with Clang's scan-build.)

scan-build was reporting NULL dereferences

Probably a bad merge from the 1.3 branch

- avoids dependency on snprintf
- allows using "smtps" instead of "456" if desired

Caused issues in threading situations

When someone defines MBEDTLS_MD5_ALT for example, the init function may need
to do more that just zeroizing the context

Allow for future support of decipherOnly and encipherOnly. Some work will be
required to ensure we still write only one byte when only one is needed.

After all it looks like those won't become standard.

Some toolchains do not have inttypes.h, and we only need stdint.h which is a
subset of it.

Introduced when moving from gmtime_r() to gmtime().
Found with fbinfer.

Hopefully makes it easier on static analyzers

Found with fbinfer.

Due to the recent change about entropy sources strength, it is no longer
acceptable to just disable the platform source. So, instead "fix" it so that
it is clear to MemSan that memory is initialized.

I tried __attribute__((no_sanitize_memory)) and MemSan's blacklist file, but
couldn't seem to get them to work.

* set_hs:
  Add tests for mbedtls_set_hs_ca_chain()
  Add tests for mbedtls_ssl_set_hs_authmode()
  Add support for SNI CA and authmode in ssl_server2
  Add mbedtls_ssl_set_hs_authmode

While at it, fix the following:
- on server with RSA_PSK, we don't want to set flags (client auth happens via
  the PSK, no cert is expected).
- use safer tests (eg == OPTIONAL vs != REQUIRED)

* snprintf:
  Rationalize other snprintf() uses
  Rationalize snprintf() usage in X.509 modules
  Add tests for snprintf
  Include fixed snprintf for Windows in platform.c

- Added in each tests program to be sure they are run (putting them in a test
  suite/function specific to the platform layer would cause them to be skipped
when PLATFORM_C is not defined).
- Platforms have already moved from a standard to a broken snprintf in the
  past [1], so make sure to catch that if it ever happens again.

[1]: http://sourceforge.net/p/mingw-w64/mailman/message/31241434/

Use _WIN32 to detect it rather that _MSC_VER as it turns out MSYS2 uses the
broken MS version by default too.

- no HAVE_TIME -> no HAVE_TIME_DATE
- check if --help works, so that the test is cleanly skipped if license server
  cannot be reached

The main goal is, we want and error if cycle counter is the only source.