Ensure clean build, no warnings on string operations using GCC 9.3.0.

Some string ops were genuinely incorrect (particularly in UT) but some
were perfectly OK and handled correctly per the C spec.  In particular
the new "rules" that GCC9 warns about make the strncat library function
(and some others) somewhat off-limits even if used correctly.

The CFE_SB_GetMsgId/SetMsgId functions, among others, were implemented
as replicas of the actual FSW code.  This creates a dependency on the
actual definition of MsgId used by the mission.

This makes the stub and actual stub.  Stubs not actually
read/write to the message in any way, they just manipulate a
local (stored in the UT framework) out-of-band buffer to hold
the metadata about the message.

This revealed a few other minor issues in test cases where they
were depending on values sitting in globals (fixed).

Move exception handling to a PSP function.  In this approach
the CFE only logs data after the event as a background job.

Replaces the CFE_ES_ProcessCoreException with a simple notification
that causes the ES background job to run, which in turn polls the
PSP for logged exceptions and writes entries to the ES ER log.

Both the PSP execption scan and the ER log file dump are converted
to background jobs.

Correct items flagged as warnings in documenation build, and
remove now-unused definition in sample platform config

Should be CFE_MISSION_SPACECRAFT_ID

The "CFE_SB_CmdHdr_t" and "CFE_SB_TlmHdr_t" types were not defined
such that they would have compatible alignment with (and thereby allow
safe casting to/from) a CFE_SB_Msg_t type.

This changes the definition to be a union so that the types are
aligned correctly.

Update the CFE_ES_ShellTlm_t, CFE_TIME_ToneSignalCmd_t, and
CFE_TIME_FakeToneCmd_t to use the CFE_SB_TlmHdr_t/CFE_SB_CmdHdr_t
types to define the buffer, rather than a uint8 array.

This should not change the size, as it was already defined using
sizeof() this structure, but it will make it aligned correctly
which resolves the compiler warning.

Note that all CMD/TLM should really be defined this way, but
this only selectively changes the places that were actually
generating a compiler warning about this.  There is a risk
that padding will be added, but this change should not change
the padding or size of messages in 32-bit builds.

Fix #637, Remove old CFE_ARINC653-related tests

Fix #484, Deprecate shell output command

Fix #641, string operations on GCC9

Fix #523, SB Subscription report control on separate MID

Fix #551, Remove duplicate CFE_TIME_Local1HzISR prototype

Fix #676, reference to deprecated CFE_SPACECRAFT_ID

Update the CFE_ES_ShellTlm_t, CFE_TIME_ToneSignalCmd_t, and
CFE_TIME_FakeToneCmd_t to use the CFE_SB_TlmHdr_t/CFE_SB_CmdHdr_t
types to define the buffer, rather than a uint8 array.

This should not change the size, as it was already defined using
sizeof() this structure, but it will make it aligned correctly
which resolves the compiler warning.

Note that all CMD/TLM should really be defined this way, but
this only selectively changes the places that were actually
generating a compiler warning about this.  There is a risk
that padding will be added, but this change should not change
the padding or size of messages in 32-bit builds.

Fix #411, rework exception handling in CFE

Fix #666, alignment of CMD/TLM message definitions

Fix #618, stubs must not depend on real msgid implementation

Fix #661, TBL events update

Fix #519, Deprecate CFS/FS time conversion APIs

Deletes now unused CFE_ES_CountObjectCallback
and CFE_ES_ListResourcesDebug.

Removes untraced capabilities to decompress libraries or
apps on load.  For non-startup apps, FS can be used.  For
startup, recommendation is to compress/decompress as part
of boot/startup vs individual applications/libs at load.

Also fixes #291, by deprecating decompress it can be moved
to the FS app.

Also makes #135 N/A, static code analysis issues in decompress

Fix #639, Remove ES_APP_DEBUG functionality

Fix #399, Deprecate decompress

Fix #559, Resolve doxygen warnings

For CFE_SB_SendMsg this was still reading the length directly
from the header.  It needs to use the metadata value instead.

Fix #491, Add UT macros