Fix #1020, refactor SB buffer descriptor object

[jphickey]

Describe the contribution
Refactor the SB buffer descriptor object (CFE_SB_BufferD_t) and simplify the zero-copy buffer paradigm.

Combine the zero-copy and the normal CFE buffer descriptor into a single unified CFE_SB_BufferD_t object. This cleans up a bunch of extra logic related to zero-copy buffers, including the extra descriptor object. The result is a simpler zero-copy design that is much less different from the standard (non-zero-copy) message path.

All message descriptor objects are now tracked in a list by SB, not just the zero-copy descriptors (for consistency - if any buffers need to be tracked, they should all be tracked).

This notably puts the buffer content as a member within the descriptor, rather than a calculated pointer, so it will be aligned properly.

Fixes #1020

Testing performed
Build and run all unit tests
Sanity check CFE

Expected behavior changes
All changes are internal to SB. This does not affect API or behavior of any existing APIs (but see note)

It also ensures that zero-copy buffers (and the associated CFE_SB_TransmitBuffer API) behave as similarly as possible to the normal CFE_SB_TransmitMsg API. Notably this corrects a minor issue where the MsgSendErrorCounter would get incremented if there were no subscribers, but only in the zero copy API.

System(s) tested on
Ubuntu 20.04

Additional context
This does not change public API in any way. However now there is no extra descriptor for the "zero copy" buffers - they are all just buffer descriptors, regardless of whether they were allocated on the fly or pre-allocated by the application.

This means that the CFE_SB_ZeroCopyHandle_t value that is output from CFE_SB_ZeroCopyGetPtr() and others is now largely redundant. It is equal to the buffer descriptor address, but it can be easily determined from the buffer content pointer too, so it is really not necessary to pass around two values for zero copy buffers. This can be simplified in a future change - but that will change the public API.

Contributor Info - All information REQUIRED for consideration of pull request
Joseph Hickey, Vantage Systems, Inc.

[jphickey]

Like others this will need a rebase when the next mainline is done, actual change is in commit 365c095

[jbohren-hbr]

Good to see this updated. This line in the current main branch increases alignment requirements on ARM32 when going from the uint8* to CFE_SB_Buffer_t*.

[skliper]

@jphickey can you rebase this one? Blocking tests on vxworks...

[astrogeco]

CCB 2021-02-17 APPROVED

• Adds to CFE_SB_BUFFER_t
• Removes CFE_SB_ZeroCopyD_t
• Consolidates information about buffer and descriptor without distinction between a "zero copy" and a "regular" one

[jphickey]

Rebased to the new main branch....

Edit: looks like a new warning popped up, so not quite ready yet - will take a look.