Fix CL_INVALID_CONTEXT command-buffer error definitions

[EwanC]

See issue #1147 documenting that the error specification for CL_INVALID_CONTEXT doesn't take into account the variation when
cl_khr_command_buffer_multi_device is enabled.

Doing this change also picked up that the error wording for clCommandSVMMemcpyKHR and clCommandSVMMemFillKHR referenced the kernel parameter which doesn't exist.

Closes #1147

[bashbaug]

Thanks! LGTM.

[bashbaug]

Could of follow-on questions:

• Does this PR fix Function clCommandBarrierWithWaitListKHR error code CL_INVALID_CONTEXT #1147? If so, note that it's not setup to close the issue when it's merged.
• Should we merge this sooner rather than later to unblock tests (Added negative tests for clCommandBarrierWithWaitListKHR OpenCL-CTS#1937) or can this wait another week or so?

Thanks!

[EwanC]

Does this PR fix Function clCommandBarrierWithWaitListKHR error code CL_INVALID_CONTEXT #1147? If so, note that it's not setup to close the issue when it's merged.

That's the intention, i've added "Closes #1147" to the PR description, which should be one of the magic words for GitHub to do this automatically.

Should we merge this sooner rather than later to unblock tests (Added negative tests for clCommandBarrierWithWaitListKHR OpenCL-CTS#1937) or can this wait another week or so?

I think this PR can wait because it's specifying relatively edge case behaviour, so I don't think waiting a week or so will affect anyone. The negative testing PRs already assume the behaviour in this PR and only stress the wording outside of the multi-device guards, this is because we've yet to do any cl_khr_command_buffer_multi_device testing so introducing it as part of negative testing even without this spec change would be a fair amount of scope increase to the task.

[bashbaug]

Merging as discussed in the July 16th teleconference.