commonfns&vectors&basic:initialize the output buffer to zeros

[banan328]

To ensure bit exactness when reading output buffers using clEnqueueReadBuffer, it's crucial to initialize all output buffers to zero. This ensures they start with a predictable state, preventing potential undefined behavior. The reason for this initialization is that in all tests, the size of the buffers exceeds the number of work items in the kernel. Without initialization, the last portion of the output buffer might contain unpredictable data ("junk").

[aharon-abramson]

In our testing environment, we check the bit-exactness of output buffers between our compiler and a reference one. Since our framework isn't aware of what part of the output buffer is actually written by a certain kernel, it compares the entire buffer; if that buffer isn't properly initialized and the kernel doesn't write all of it, some of its content is undefined.

[bashbaug]

Discussed in the June 4th teleconference (I think, may have been the week prior). Concerns are that this "buffers must be initialized" property is not enforced by the current CTS tests, so even if we merge this PR there is no guarantee that these buffers will continue to be initialized, nor is there a way to have confidence that there are not other tests that would need to be updated to initialize buffers.

Given these concerns, would a better solution be for whatever testing framework that is requiring bit exactness to do its own buffer intialization, vs. doing the buffer intialization explicitly in the tests?

Removing "focused review" while evaluating whether these changes are required.